Preliminary Assessment of Volcanic and HydrothermalHazards in Yellowstone National Park and Vicinity

By Robert L. Christiansen, Jacob B. Lowenstern, Robert B. Smith, Henry Heasler, Lisa A. Morgan, ManuelNathenson, Larry G. Mastin, L. J. Patrick Muffler, and Joel E. Robinson

Open-file Report 2007–1071

U.S. Department of the InteriorU.S. Geological Survey

ii

U.S. Department of the InteriorDIRK KEMPTHORNE, Secretary

U.S. Geological SurveyMark D. Myers, Director

U.S. Geological Survey, Reston, Virginia 2007

For product and ordering information:World Wide Web: http://www.usgs.gov/pubprodTelephone: 1-888-ASK-USGS

For more information on the USGS—the Federal source for science about the Earth,its natural and living resources, natural hazards, and the environment:World Wide Web: http://www.usgs.govTelephone: 1-888-ASK-USGS

Suggested citation:Christiansen, R. L., Lowenstern, J. B., Smith, R. B., Heasler, H., Morgan, L. A., Nathenson, M., Mastin, L. G.,Muffler, L. J. P., and Robinson, J. E., 2007, Preliminary assessment of volcanic and hydrothermal hazards inYellowstone National Park and vicinity: U.S. Geological Survey Open-file Report 2007-1071, 94 p.

Any use of trade, product, or firm names is for descriptive purposes only and does not implyendorsement by the U.S. Government.

Although this report is in the public domain, permission must be secured from the individualcopyright owners to reproduce any copyrighted material contained within this report.

iii

ContentsContents................................................................................................................................................................ iiiBrief Summary.......................................................................................................................................................1Introduction ...........................................................................................................................................................4Factors considered in this assessment................................................................................................................5Current State of the Yellowstone system.............................................................................................................5

Geologic background ........................................................................................................................................5Contemporary activity........................................................................................................................................8

Seismicity .......................................................................................................................................................8Crustal deformation...................................................................................................................................... 11Hydrothermal and gas activity..................................................................................................................... 12

Pressures, temperatures, and fluids in geothermal systems ................................................................. 13Mechanisms of hydrothermal explosion ................................................................................................. 14Hydrothermal explosions in Yellowstone ................................................................................................ 15Factors contributing to hydrothermal explosions ................................................................................... 16Toxic gases ............................................................................................................................................... 17

The Hazards......................................................................................................................................................... 17Volcanic-eruption hazards .............................................................................................................................. 18

Basaltic eruptions ........................................................................................................................................ 18Large rhyolitic eruptions .............................................................................................................................. 20Small rhyolitic eruptions .............................................................................................................................. 25Large caldera-forming eruption................................................................................................................... 26

Hydrothermal-explosion hazards .................................................................................................................... 29How often do they occur?............................................................................................................................ 29Potential effects ........................................................................................................................................... 30Precursory signals ....................................................................................................................................... 31Where are hydrothermal explosions most likely to occur?........................................................................ 31Seasonal and long-term effects on hydrothermal explosions ................................................................... 33Hazard mitigation ......................................................................................................................................... 33

Gas-emission hazards ..................................................................................................................................... 34Relevant examples of toxic volcanic or hydrothermal gas hazards .......................................................... 35Hazard mitigation ......................................................................................................................................... 36

Conclusions ......................................................................................................................................................... 36Acknowledgments............................................................................................................................................... 37References Cited................................................................................................................................................. 37Appendix 1. Description of representative historic hydrothermal explosions ................................................. 86

Porkchop Spring/Geyser ................................................................................................................................. 86Excelsior Geyser .............................................................................................................................................. 86West Nymph Creek Thermal Area .................................................................................................................. 87Black Opal/Wall Pool and Sapphire Pool ....................................................................................................... 87Historic hydrothermal explosions elsewhere................................................................................................. 87

Appendix 2. Description of large prehistoric hydrothermal eruption sites at Yellowstone ............................. 88Pocket Basin .................................................................................................................................................... 88Mary Bay.......................................................................................................................................................... 88Elliott’s crater ................................................................................................................................................... 89Evil Twin explosion crater ............................................................................................................................... 89Frank Island explosion crater.......................................................................................................................... 89Indian Pond ...................................................................................................................................................... 89Turbid Lake....................................................................................................................................................... 90

iv

Appendix 3. Probabilities of episodic volcanic eruptions and application to the young intracaldera volcanichistory of Yellowstone......................................................................................................................................... 91

1

Preliminary Assessment of Volcanic andHydrothermal Hazards in Yellowstone National Parkand Vicinity

By Robert L. Christiansen1, Jacob B. Lowenstern1, Robert B. Smith2, Henry Heasler3, Lisa A.Morgan1, Manuel Nathenson1, Larry G. Mastin1, L. J. Patrick Muffler1, and Joel E. Robinson1

Brief SummaryPossible future violent events in the active hydrothermal, magmatic, and tectonic system of

Yellowstone National Park pose potential hazards to park visitors and infrastructure. Most of thenational park and vicinity are sparsely populated, but significant numbers of people as well as parkresources could nevertheless be at risk from these hazards. Depending on the nature and magnitudeof a particular hazardous event and the particular time and season when it might occur, 70,000 tomore than 100,000 persons could be affected; the most violent events could affect a broader regionor even continent-wide areas. This assessment of such hazards is presented both as a guide forfuture activities of the Yellowstone Volcano Observatory (YVO) and to aid appropriate responseplanning by the National Park Service and surrounding agencies and communities. Although theassessment is presented here in some technical detail, this summary is intended to beunderstandable to non-scientists. The principal conclusions also will be made available in otherforms, more accessible to general readers.

The Yellowstone Plateau was built by one of Earth's largest young volcanic systems, havingepisodically erupted great volumes of both lava and explosively ejected pumiceous ash for morethan 2 million years. These eruptive materials are products of two compositional types ofsubsurface magma: basaltic magma is relatively fluid and, in this setting, generally produces smallto moderate volumes of lava in relatively brief eruptions; rhyolitic magma is more viscous andeither can erupt effusively to produce small to large volumes of lava or explosively to producecoarse pumice and finer ash. The three largest Yellowstone eruptions produced blanketing depositsof rhyolitic ash so hot that they welded into sheets of dense rock covering large areas, extendingbeyond the national park. Each of them also produced a rain of ash that spread over much ofwestern and central North America and beyond; these ash deposits are greater than 2 m thick neartheir eruptive sources and as much as a meter thick in surrounding areas. Each of these threeeruptions produced a caldera, or deep crater-like depression, tens of kilometers wide, formed bycollapse of the ground surface into a partly emptied subterranean magma chamber. The latest ofthese three great eruptions formed the Yellowstone caldera. Renewed rhyolitic magma influxbeneath the Yellowstone caldera in central Yellowstone National Park uplifted parts of the calderafloor and produced voluminous intracaldera lavas, the youngest of which extruded in a series oferuptive episodes about 164,000, 152,000, 114,000, 102,000, and 72,000 years ago. During thesame span of time, generally smaller flows of both basalt and rhyolite have erupted in several areasoutside the Yellowstone caldera: (1) northwest of the caldera, (2) near the southern boundary of the 1 U.S. Geological Survey2 University of Utah3 National Park Service

2

park, (3) in the basin of Island Park, west of Yellowstone National Park, and (4) especially in thesouthern part of a corridor between Norris Geyser Basin and Mammoth Hot Springs.

Disruption of the Earth’s surface by faulting and regional uplift characterize the geologicframework of Yellowstone Plateau volcanism. Some of the regional faults that bound the mountainranges around the Yellowstone Plateau are capable of producing large-magnitude (M>7)earthquakes. In contrast, faults within the caldera are mainly small, produce smaller-magnitude(M 6.8), relatively shallow earthquakes, and reflect strains in the Earth’s crust associated withmagmatic intrusion and hydrothermal activity. Swarms of generally small earthquakes occurringwithin localized areas over restricted periods of time characterize much of the earthquake activitywithin and adjacent to the Yellowstone caldera. This seismicity is monitored by a network ofseismographs within and adjacent to the park and is recorded and processed nearly in real time atthe University of Utah as part of the YVO program and archived as a contribution to the U.S.Geological Survey (USGS) Advanced National Seismic System database.

Leveling surveys, satellite-based measurements, and geologic studies of former shorelinesof Yellowstone Lake all show that the entire area of the Yellowstone caldera and a seismicallyactive zone to the northwest undergo episodes of ground uplift and subsidence, sometimesencompassing the entire caldera and sometimes in more local and complex patterns of both upliftand subsidence. Such deformation in the Yellowstone region is monitored by YVO mainly througha network of continuously recorded Global Positioning System (GPS) receivers recorded at theUniversity of Utah. The GPS data are incorporated as part of the National Science Foundation’sEarthScope Plate Boundary Observatory, archived at UNAVCO and available through the YVOwebsite (http://volcanoes.usgs.gov/yvo/index.html).

The active hydrothermal system of Yellowstone National Park is one of the largest onEarth. Although accidents involving hot water injure Yellowstone visitors from time to time,conformance with normal Park Service procedures and regulations would ordinarily be sufficient toprevent most of them. By contrast, a commonly recurring, more acute hazard at Yellowstone is theexplosive ejection of steam, water, and rock with no associated volcanism. These hydrothermalexplosions are caused by hot subsurface waters that flash to steam, breaking the overlying rocksthat confine them and ejecting the debris to form a crater. It is generally not clear just what triggersthese events, but possible triggers include strong local earthquakes, seasonal or long-term declinesin ground water levels, and changes in the underground distribution of heat. Many hydrothermalexplosions have few if any premonitory indications.

At least 26 hydrothermal explosions have been documented in the 126-year historic recordof the national park, and others undoubtedly escaped observation. Since the Yellowstone Plateauwas last glaciated, ending about 16,000 years ago, at least 18 large hydrothermal explosions haveformed craters wider than 100 m. Conservatively, at least one rock-hurling explosion every twoyears is estimated to occur at Yellowstone, but because most of these events are small and usuallyoccur when few visitors are present, the likelihood of harm to park visitors is relatively small. Theaverage recurrence of an explosion that could cause personal injury is probably between 10 and 100years. Average recurrence time of an explosion large enough to produce a 100-m-diameter crater isprobably about 200 years, but such an event could expel rocks and other hot debris more than 2 kmfrom the explosion site. Most hydrothermal-explosion craters at Yellowstone are in the FireholeRiver geyser basins, beneath or around Yellowstone Lake, and in the southern part of the Norris-Mammoth Corridor.

In addition to hydrothermal explosions, toxic gases—mainly carbon dioxide and hydrogensulfide—pose hazards. Concentrations of these gases in the atmosphere are generally low atYellowstone, but they can build up in confined areas such as valleys, caves, and tunnels, especiallyduring windless conditions. Most areas with toxic-gas hazards can be kept off-limits to people, butgas emissions should continue to be monitored.

3

No volcanic eruption has occurred in Yellowstone National Park or vicinity in the last70,000 years or more. Nevertheless, several types of volcanic eruption hazards are possible in thefuture.

Basaltic lavas have erupted around the margins of the Yellowstone Plateau volcanic fieldthroughout its evolution. These relatively low-viscosity lavas generally erupt rapidly, mosteruptions lasting no more than a few weeks to a few months though the largest flow fields mayaccumulate in multiple eruptions lasting months to years. The average period between basalticeruptions in the Yellowstone region since formation of the Yellowstone caldera has been about16,000 years. The most likely location of a future basaltic eruption is within the basin of IslandPark, west of Yellowstone, but basalts could erupt anywhere in a 40-km-wide band around thecaldera. Future basaltic eruptions could cover several square kilometers with lava up to tens ofmeters thick. Basaltic ash and cinders also might blanket hundreds of square kilometers to depthsof a few meters to a few centimeters, and if a vent emerged beneath water or saturated ground,more explosive eruptions could cause significant destruction, such as blasting down trees orstructures.

Large rhyolitic lava flows, many having volumes greater than 10 km3, have erupted withinthe Yellowstone caldera during the past 170,000 years. Initially these larger eruptions werepreceded by explosively ejected pumice and ash. In a similar future eruption, ejecta could burybroad areas, locally to many meters. Subsequent lava extrusion could last for years, covering areasas great as 350-400 km2 to thicknesses of tens or hundreds of meters and volumes of 5 to more than50 km3. Because such voluminous rhyolitic lava eruptions have not been observed anywhere inhistorical time, it is uncertain how long such an event might continue; extrusion might be orders ofmagnitude faster than for smaller flows. The probability of a future large intracaldera rhyoliticeruption is difficult to estimate. Available data suggest a highly episodic behavior of past eruptionsof this sort, periods of a few thousand years characterized by numerous eruptions being separatedby longer intervals of about 12,000 to 38,000 years without eruption. One statistical measure oferuption probabilities based on this episodic behavior suggests an average recurrence of 20,000years. The fact that no such eruption has occurred for more than 70,000 years may mean thatinsufficient eruptible magma remains beneath the Yellowstone caldera to produce another large-volume lava flow.

Small rhyolitic lava flows postdating the Yellowstone caldera have erupted mainly north ofthe caldera, but one such flow also lies near the South Entrance to the park. Two distinct types ofprimary hazards might be associated with small rhyolitic eruptions at Yellowstone. Just as forlarger rhyolitic lava eruptions, initial venting almost certainly would explosively eject rhyoliticpumice; the coarser fragments would fall back close to the vent, but finer pumiceous ash wouldenter the atmosphere and fall downwind for many kilometers. Structures, power lines, etc. could bedamaged by ash loading, especially if eruption were accompanied by heavy rain. The initialexplosive eruptions could last a few hours to several weeks and be followed by viscous extrusion ofrhyolitic lava, covering several square kilometers to tens of meters thick; lava could continue toextrude for many months or even years. Viscous rhyolitic lava would advance much more slowlythan a basaltic flow; most affected facilities could be safely evacuated and perhaps relocated. Theaverage recurrence period of small extracaldera rhyolitic eruptions in the Yellowstone Plateauvolcanic field is about 50,000 years.

In addition to the primary hazards posed by any future eruption of basalt or a small or largerhyolitic lava eruption, important possible secondary consequences include wildland fires, debrisflows, and floods triggered by the displacement of surface drainages by lava.

Systematic seismic, deformation, and hydrothermal monitoring by YVO is likely to provideindicators of any impending volcanic eruptive activity in Yellowstone National Park. Premonitoryevents detected by such monitoring might include multiple shallow earthquake swarms of

4

increasing frequency and intensity, the ground vibrations called volcanic tremor, localized uplift ofthe surface, ground cracks, and anomalous gas emissions.

Of all the possible hazards from a future volcanic eruption in the Yellowstone region, by farthe least likely would be another explosive caldera-forming eruption of great volumes of rhyoliticash. Abundant evidence indicates that hot magma continues to exist beneath Yellowstone, but it isuncertain how much of it remains liquid, how well the liquid is interconnected, and thus how muchremains eruptible. Any eruption of sufficient volume to form a new caldera probably would occuronly from within the present Yellowstone caldera, and the history of postcaldera rhyolitic eruptionsstrongly suggests that the subcaldera magma chamber is now a largely crystallized mush. Theprobability of another major caldera-forming Yellowstone eruption, in the absence of strongpremonitory indications of major magmatic intrusion and degassing beneath a large area of thecaldera, can be considered to be below the threshold of useful calculation.

IntroductionYellowstone National Park, justly famous for its unmatched geysers, diverse wildlife, and

uniquely preserved ecologic communities, also encompasses one of Earth’s largest systems ofvolcanic, seismic, and hydrothermal activity. In recognition of the importance of this active Earthsystem, officials of the U.S. Geological Survey (USGS), Yellowstone National Park, and theUniversity of Utah in May of 2001 jointly established the Yellowstone Volcano Observatory(YVO). The stated objectives of this new observatory are: (1) to provide monitoring that enablesreliable and timely warnings of possible renewed volcanism and related hazards in the Yellowstoneregion, (2) to notify National Park and other local officials and the public of any significant localseismic or volcanic events, (3) to improve scientific understanding of the fundamental tectonic andmagmatic processes that create the Park’s ongoing seismicity, surface deformation, andhydrothermal activity, (4) to assess the long-term potential hazards that volcanism, seismicity, andexplosive or other convulsive hydrothermal activity might pose to the park and its surroundings, (5)to communicate as effectively as possible to responsible authorities and the public the results ofscientific monitoring and hazard-assessment activities, and (6) to improve coordination andcooperation among the three institutions responsible for YVO. The observatory is built upon asubstantial base of previous cooperative work among these institutions and seeks to assure a solidlong-term basis for the continuity and improvement of scientific monitoring of the Yellowstonemagmatic-tectonic-hydrothermal system.

The hazard assessment presented in this report is intended to help guide future activities ofthe observatory as well as to provide a basis for appropriate management actions by the NationalPark Service and other agencies in the Yellowstone area in the event of any future hazardous eventsthat might result from activity of the Yellowstone system. The assessment is part of an ongoingthree-part process that was set in motion at the outset of YVO’s work. The first part of the processwas a comprehensive review of basic scientific knowledge of Yellowstone’s magmatic-tectonic-hydrothermal system. Such a review, of course, is never final and must be continually reexaminedand revised, particularly in the light of current monitoring data. The second step of the process isassessment of the relevant hazards. The current hazards assessment is intended to be fullydocumented and scientifically defensible; because this necessarily entails a degree of scientificrigor and technical documentation not readily understandable by general readers or to all concernedindividuals, additional reports more suitable for non-scientist readers also are part of the ongoingassessment activity. One general-interest publication on Yellowstone’s volcano, seismic, andhydrothermal hazards has already been published (Lowenstern and others, 2005). The final step ofthe process will be a response plan, based upon the conclusions of this assessment, by theresponsible Yellowstone National Park officials, in cooperation with other appropriate agencies andwith the assistance of observatory scientists.

5

It is worth noting here that, despite the fact that most of Yellowstone National Park and itsnearby surroundings are sparsely populated, many thousands of people as well as National Parkresources could be at risk from the types of hazards considered in this assessment. About 3 millionpeople visit Yellowstone each year, principally during the three summer months. A similar numberof people visit adjacent Grand Teton National Park (although some of these are the same visitors).Including residents of the surrounding communities of Wyoming, Idaho, and Montana, the dailypopulation exposure to these hazards during the summer months could average between 70,000 andmore than 100,000 persons. This exposure produces a largely unappreciated level of risk,comparable to that of other areas having considerably larger resident populations. Furthermore, thelargest of these hazards, although they have the lowest probabilities of occurrence, could affectmuch of western and central North America. Indirect effects, especially on climate, could beglobal.

Factors considered in this assessmentIn order to clarify the intended scope of this report, it is important to state at the outset just

what factors are considered explicitly in this assessment. Hazards that might result from ongoingor future activity of the Yellowstone magmatic-tectonic-hydrothermal system, particularly volcaniceruptions, earthquakes, and hydrothermal eruptions, are the focus of this discussion. Certain othergeologic hazards, although likely to be of important concern to National Park Service managers inthe future, are not included within its scope. Examples of the latter would include landslides,debris flows, or floods, except insofar as they are considered here as possible consequentialsecondary hazards that might result from major volcanic, seismic, or hydrothermal activity.

The area considered in this hazard assessment is primarily Yellowstone National Park (fig.1). Nevertheless, because most of the hazards considered here could have significant effects onadjacent communities, the Park boundary does not constitute an absolute limit in our analysis.

The strategy is to establish an integrated view of the sources of relevant hazards and somescenarios for how they might develop in time. This approach diverges from that of many previousUSGS volcano-hazard assessments in not emphasizing a catalogue of individual hazardousprocesses and zonation maps for each of them. Rather, the emphasis is on (1) what kinds ofmonitoring data might be of immediate concern to National Park Service managers or might beconsidered premonitory to hazardous events, (2) the probabilities of recurring volcanic and acutehydrothermal events, and (3) how multiple events might be related to one another. Although thepresent open-file version of this report does not analyze earthquake hazards explicitly, aforthcoming revision for more formal publication will include seismic hazards.

Current State of the Yellowstone systemAssessment of possible future activity in Yellowstone starts from an analysis of the current

state of the magmatic-tectonic-hydrothermal system. We first review the geologic framework andfollow with information on current activity and monitoring of the system.

Geologic backgroundThe Yellowstone Plateau volcanic field of Wyoming, Idaho, and Montana (fig. 2) is one of

Earth's largest young volcanic systems, having erupted extraordinarily voluminous rhyolitesepisodically over a little more than 2 million years (Christiansen, 1984; 2001). Its three largesteruptions deposited sheets of mainly welded ash-flow tuffs regionally as well as coeval ash-falllayers that fell over much of western and central North America (fig. 3). The volcanic field is thecurrent expression of a major sublithospheric mantle source that generates a melting anomaly that

6

has propagated northeastward at least 300 km relative to the North America craton since its initialvolcanism at 17-14 Ma (fig. 4). Researchers debate the nature of this melting source—whether, forexample, it represents a convective thermal plume from the base of the mantle or from the mantletransition zone, an entirely upper-mantle response to plate-tectonic processes, or some othermechanism (e.g., Pierce and Morgan, 1992; Smith and Braile, 1993, 1994; Pierce and others, 2000;Christiansen and others, 2002; Camp and Ross, 2004; Waite and others, 2006). These variedhypotheses are significant to interpretation of regional geophysical data but have only an indirectbearing on the present analysis of volcanic hazards.

The volcanic field has evolved episodically in three cycles of rhyolitic activity. Each cycleculminated in the rapid eruption of voluminous rhyolitic ash flows—hundreds to thousands ofcubic kilometers—and consequent catastrophic subsidence of the source areas to form largecalderas. Each climactic ash-flow eruption was preceded by a period of magmatic intrusion andintermittent rhyolitic lava eruptions and was followed by a period of partial filling of the calderaswith rhyolitic lavas. During each of these cycles, basaltic lavas erupted around the margins of theactive rhyolitic source area but not within it. About a million years after their rhyolitic activity,basalts finally erupted through the source areas of the first two cycles, but no basalts have yeterupted within the youngest, the Yellowstone caldera.

Collectively, the voluminous rhyolitic ash-flow tuffs of the volcanic field are knownstratigraphically as the Yellowstone Group (Christiansen and Blank, 1972). The oldest and largestof the caldera-forming eruptions produced the Huckleberry Ridge Tuff at 2.059±0.004 Ma(Lanphere and others, 2002), covering an area of more than 15,000 km2 with ash flows having acumulative volume of at least 2,450 km3. The resulting caldera (fig. 5, purple line) spanned nearlythe entire width of the Yellowstone Plateau volcanic field (Christiansen, 1979, 2001). The MesaFalls Tuff of 1.285±0.004 Ma (Lanphere and others, 2002), smallest of the three ash-flow sheets,erupted in the Island Park area west of the Yellowstone Plateau (Christiansen, 1982, 2001) and isnow exposed mainly in that vicinity, adjacent to its source caldera (fig. 5, blue line). The MesaFalls probably initially covered an area of more than 2,700 km2 and had an eruptive volume ofmore than 280 km3. The youngest of the major caldera-forming eruptions at 0.639±0.002 Ma(Lanphere and others, 2002) produced the Lava Creek Tuff and the Yellowstone caldera, now thecentral feature of the volcanic field (Christiansen, 1984). Two distinct parts of the Lava Creek Tufferupted from separate segments of the caldera (Christiansen, 1979, 2001) but form a singlecompound cooling unit that initially covered at least 7,300 km2 and had an eruptive volume of atleast 1,000 km3.

Precaldera rhyolitic lavas of the third volcanic cycle erupted from a growing ring-fracturesystem in the area that subsequently became the Yellowstone caldera. Eight such precalderarhyolite flows are known, yielding isotopic ages from 1.22±0.01 to 0.609±0.006 Ma (table 1; seenote on the table regarding the accuracy of cited ages). Additional precaldera rhyolitic lava flowsmay now lie entirely buried within the Yellowstone caldera. The precaldera lavas, named theMount Jackson Rhyolite and Lewis Canyon Rhyolite, and the growing ring-fracture system fromwhich they extruded (fig. 6) are interpreted as indicating magmatic intrusion and the growth of alarge rhyolitic magma chamber in the shallow crust during a period of nearly 600,000 years beneaththe area that would later erupt catastrophically to produce the Lava Creek Tuff and subside to formthe Yellowstone caldera (fig. 7).

Shortly following the climactic caldera-forming Lava Creek eruption, rhyolitic magmaagain intruded the subcaldera region, uplifting segments of the caldera floor in a pair of resurgentdomes bounded by two inner ring-fracture segments enclosed by an outer ring-fracture zoneencompassing nearly the entire Yellowstone caldera (Christiansen, 2001). The western of theseresurgent domes is the Mallard Lake dome (fig. 7, ML); the eastern is the Sour Creek dome (fig. 7,SC). Subsequently, additional rhyolitic lavas have erupted within the caldera during several

7

volcanic episodes. These rhyolites, mainly constituting large lava flows, are includedstratigraphically in the Plateau Rhyolite. The oldest of the Plateau Rhyolite flows, the Upper BasinMember, erupted within the caldera from around the inner ring-fracture zones that bound the tworesurgent domes (fig. 7). These oldest postcaldera rhyolites yield isotopic ages that range between516±7 and 479±10 ka (Gansecki and others, 1996) though there is a reasonable possibility that theiractual eruptive ages were even closer to that of the ~639-ka Lava Creek Tuff (see Lanphere andothers, 2002). Buried rhyolitic lavas identified in recent high-resolution sonar and seismic-reflection mapping of Yellowstone Lake (Morgan and others, 2003b) and found as lithic clasts inpostglacial hydrothermal-explosion deposits south of the Sour Creek dome and in northernYellowstone Lake yield a 40Ar/39Ar age of 600±20 ka (Morgan and Shanks, 2005). Additionalpostcaldera lava, probably as young as 275±11 ka is known (table 1). Other lavas may well haveextruded within the caldera during the time before about 170 ka but remain buried beneath youngerlavas and sediments.

Younger postcaldera rhyolitic volcanism dates from about 170 ka (Christiansen, 2001).This activity probably began with extrusion of rhyolitic lava onto the Mallard Lake resurgent domein the western part of the Yellowstone caldera (the Mallard Lake flow) followed by renewed upliftof that dome. Following those events, as much as several hundred cubic kilometers of rhyoliticlava (figs. 8 and 9), named the Central Plateau Member of the Plateau Rhyolite, has nearly filledthe Yellowstone caldera from eruptive vents along two linear northwest-trending zones that projectacross the caldera from neighboring tectonic fault zones. The western of these two zones extendsfrom the Teton fault zone (see fig. 11) to the tectonic basin of West Yellowstone. Lavas from thesevents form the Pitchstone and Madison Plateaus (fig. 1) and bury the western rim of theYellowstone caldera. The eastern zone extends northward across the caldera from the Sheridanfault zone (see fig. 11) to the Norris Geyser Basin and erupted the lavas that form the CentralPlateau. The lava flows from this youngest intracaldera activity are typically quite large—some ofthem exceeding 20 km3—and appear to have erupted in five episodes at about 164±5, 152±3,114±2, 102±5, and 72±4 ka (table 2 and figs. 8 and 9). It is likely that rhyolitic pumice and asherupted during the opening of vents for each of these lava flows. In addition, concurrent with lavaeruptions of the Central Plateau Member, at least two significant pyroclastic eruptions occurredwithin the caldera, depositing the tuffs of Bluff Point and Cold Mountain Creek. The former wasof sufficiently large volume to have caused subsidence of its source area to form a smaller caldera(~10-km diameter) within the Yellowstone caldera, now partly preserved as West Thumb (fig. 8),the western part of Yellowstone Lake (Christiansen, 2001). It is worth noting that this relativelysmall caldera is itself as large as the well-known caldera of Crater Lake, Ore., which formed as aresult of the pyroclastic eruption of more than 50 km3 of magma (Bacon, 1983).

Volcanism postdating the Yellowstone caldera also has occurred in areas outside the calderabut within or near the boundaries of Yellowstone National Park. The individual eruptive productsof this extracaldera volcanism are generally smaller in volume than most of those within the calderabut span nearly the same range of time as the intracaldera lavas. Both rhyolites and basalts occuramong these extracaldera lavas. K/Ar and 40Ar/39Ar ages obtained on the rhyolites range between526±3 and 80±2 ka (table 1). Basaltic eruptions appear to have spanned much of the same time.Basalts and rhyolites both erupted mainly in the southern part of the Norris-Mammoth corridor, azone of faulting, volcanism, and hydrothermal activity that extends northward from the calderamargin near Norris Geyser Basin to just north of Gardiner, Mont. (figs. 10, 11). Other rhyolites andbasalts of similar ages have erupted northwest of the caldera, between the Madison River andCougar Creek, and near the southern boundary of Yellowstone National Park. Furthermore, basaltserupted from at least 17 vents in the basin of Island Park, within 35 km of the west boundary ofYellowstone National Park (fig. 10).

8

Hydrothermal activity occurs widely on and around the Yellowstone Plateau. Most of thehottest and most active areas of geysers and other near-neutral (i.e., non-acidic) hot springs occurwithin topographic basins situated above the ring-fracture zone of the Yellowstonecaldera—including Yellowstone Lake (Johnson and others, 2003; Morgan and others, 2003a;Morgan and others, 2003b)—but others occur in the Norris-Mammoth corridor north of the caldera.Other major sites of mainly acidic, sulfate-rich hot-springs and fumaroles occur in topographicallyhigher areas in and adjacent to the ring-fracture zone and the northeastern caldera rim. Smaller,less concentrated zones of hydrothermal activity are distributed widely around the YellowstoneNational Park, but most areas deemed capable of producing violent events such as hydrothermalexplosions are in the major hydrothermal areas within or adjacent to the caldera and in the Norris-Mammoth corridor.

Regional uplift and normal faulting have established the geologic framework within whichthe Yellowstone Plateau volcanic field lies (figs. 4 and 11). Faults mapped within the caldera aremainly small and seem to reflect strains associated with magmatic intrusion and, perhaps, shallowhydrothermal activity. The principal faults south of the caldera that accommodate regional tectonicextension generally trend nearly north-south and define the Teton, Sheridan, Flat Mountain, andUpper Yellowstone fault zones. Major fault zones north of the caldera include the Lamar, EastGallatin-Washburn, Hebgen, and Madison Valley fault zones, many of them trending more nearlynorthwest. A little farther west, the Centennial fault zone trends east-west. The somewhat arcuateMirror Plateau fault zone northeast of the caldera (figs. 6 and 11) appears to have accommodatedboth regional tectonic extension and displacements associated directly with the Yellowstonecaldera system (Christiansen, 2001).

Bathymetric and seismic-reflection studies have delineated several small faults beneathYellowstone Lake (Otis and others, 1977; Wold and others, 1977; Johnson and others, 2003;Morgan and others, 2003b). At least the major sublacustrine faults appear to representcontinuations of the regional tectonic trends noted above (Morgan and others, in press-b).

Thermal surveys of Yellowstone Lake (Morgan and others, 1977), together withgeochemical indicators of the heat output of the Yellowstone hydrothermal system (Fournier,1989), demonstrate that the total heat flux from the Yellowstone caldera exceeds 1500 mW/m2,more than thirty times the regional average. This great thermal flux is a direct reflection of themagmatic heat source that produces and sustains Yellowstone’s hydrothermal activity.

Contemporary activityThe tectonic-volcanic-hydrothermal system of the Yellowstone region is vigorously active.

The hydrothermal system may be the largest on Earth, and the subsurface magmatic systemcontinually deforms the ground surface on time scales of months to a few years. Regional tectonicsand the magmatic system combine to produce some of the highest levels of earthquake activity inthe conterminous U.S. outside of California.

SeismicityAlthough physiographically part of the Middle Rocky Mountains, Yellowstone lies along

the northeastern margin of the extensional basin-range tectonic region (figs. 4 and 11). Epicentersof earthquakes associated with this tectonic extension define a belt of seismicity that extends northfrom the Wasatch front in Utah to the Yellowstone Plateau, then branches to the northwest andwest into Montana and Idaho, termed the intermountain seismic belt by Smith and Sbar (1974).Earthquake epicenters define a parabolic arc around the north, east, and south sides of the easternSnake River Plain through the Yellowstone Plateau, where the seismicity is most active (Smith andArabasz, 1991; Smith and Braile, 1994) and fault displacements are greatest (Anders and others,

9

1989; Pierce and Morgan, 1992). Earthquake distributions, focal mechanisms, and GPSdeterminations of crustal strain in the Yellowstone Plateau area are all consistent with the generallyNE-SW direction of extension in the basin-range region (Waite and Smith, 2004; Puskas andothers, 2007).

Seismicity in the Yellowstone National Park region is monitored in real time by a networkof 26 seismic stations within or adjacent to the park (fig. 12), all of which are recorded at theUniversity of Utah as an integral part of the YVO program and archived in the Advanced SeismicNetwork System. Of these seismic stations, 17 are single-component short-period seismometersconnected to the network by analog telemetry. Additionally, there are three short-period 3-component seismometers with analog telemetry and six broadband 3-component seismometers withdigital telemetry (Yellowstone Volcano Observatory, 2006). Data are transmitted continuouslyfrom the seismic stations via radio signals to a Federal Aviation Administration (FAA) radar sitelocated on Sawtell Peak, Idaho, west of Yellowstone National Park, where the signals aremultiplexed onto four FAA microwave lines for retransmission to the FAA control center in SaltLake City, Utah. From there, the data are transmitted to the University of Utah's central recordinglaboratory via telephone lines. Earthquake data are processed using the USGS EarthWorm systemto produce automated real-time hypocenter determinations and emergency-response broadcasts.Refined earthquake locations, magnitudes, and focal mechanisms are produced by seismic analystsat the University of Utah Seismograph Stations from the digital data and are provided to YVOusers.

Given its average station spacing of 15-30 km and seismometer characteristics, theYellowstone network is optimally capable of detecting local earthquakes of magnitudes greaterthan about -1.0 and locating their epicenters within about 1 km and their focal depths within about0.5 km at depths below about 4 km (or about 1 km at shallower depths). Analysis of the digitalseismic data includes production of a revised catalog of Yellowstone earthquakes using new three-dimensional P-wave velocity models determined by tomographic inversion of local earthquakes(Husen and Smith, 2004) and a new magnitude scale for improved hazard assessments.

It is useful to consider seismicity of the Yellowstone region in terms of several categories ofactivity. Most Yellowstone earthquakes (fig. 13) are of small magnitude (M 3), but a few largetectonic earthquakes have affected the region, including the deadly Hebgen Lake earthquake ofAugust 17, 1959, with a magnitude of 7.5 (U.S. Geological Survey, 1964; Trimble and Smith,1975; Doser, 1985; Doser and Smith, 1989). Aftershocks of the Hebgen Lake earthquake werenumerous (Murphy and Brazee, 1964), and a high proportion of the earthquakes since then haveoccurred in the same belt of seismicity as those early aftershocks (Trimble and Smith, 1975; Smithand others, 1977; Smith and Braile, 1994). Numerous other earthquakes, including many withinthe Yellowstone caldera, are scattered more widely, are typically shallow, and commonly occur inthe form of seismic swarms. Seismicity beneath the Yellowstone caldera is generally shallower (~5km) relative to the deepest earthquakes (~20 km) on tectonic faults outside the caldera, probablyrelated to the effects of elevated crustal temperature above the magma that underlies the caldera(Miller and Smith, 1999; Waite and Smith, 2002; Husen and Smith, 2004). Thus, seismicity can bediscussed in four categories: large earthquakes on regional tectonic faults, smaller tectonicearthquakes, intracaldera earthquakes, and seismic swarms.

The 1959 Hebgen Lake earthquake had the largest magnitude of any historic earthquake inthe Rocky Mountains region (fig. 13). That earthquake produced a pair of fault scarps up to 6 mhigh and 40 km long, caused a landslide that dammed the Madison River northwest of HebgenLake about 30 km west of Yellowstone National Park, caused extensive damage to roads andbuildings within the park, and resulted in 28 deaths—all of them in the National Forest, outside thepark. The only other regional historic earthquake of roughly comparable magnitude was the M=7.3Borah Peak, Idaho, earthquake of 1983, about 150 km southwest of Yellowstone National Park

10

(Dewey, 1983; Smith and others, 1985; Crone, 1987; Richins and others, 1987). It too produced along surface scarp and resulted in local damage. Both of these earthquakes caused discernablechanges in the hydrothermal features of Yellowstone (Watson, 1961; Marler, 1964; Marler andWhite, 1975; Hutchinson, 1985). Prehistoric late Holocene scarps occur on several range frontsnear Yellowstone, including the Centennial, Madison, and Teton Ranges (pl. 1 of Pierce andMorgan, 1992), indicating the possibility of further large tectonic earthquakes in the region.

Smaller tectonic earthquakes on regional faults occur widely in areas in and adjacent toYellowstone National Park (Smith and Sbar, 1974; Doser and Smith, 1983; Waite and Smith, 2002,2004; White, 2006). Recurring seismicity of this type is common and must be expected in thefuture. Most of these earthquakes are small and cause no significant damage. Others are feltlocally, and some have the potential to cause damage to structures or other facilities.

Intracaldera earthquakes are typically small (M 3) and generally are shallower than about 6km (Miller and Smith, 1999; Husen and Smith, 2004; Waite and Smith, 2004). A few however, arelarger and can be damaging. The largest earthquake recorded within the caldera area occurredsoutheast of Norris Geyser Basin (fig. 13) in June, 1975 with a magnitude of 6.1 (Pitt and others,1979). The earthquake, felt widely within Yellowstone National Park, caused damage to roads andother visitor facilities. The area in which it was felt was more extensive to the north, outside thecaldera, than within the caldera, probably because of the attenuating effects of elevated crustaltemperatures beneath the caldera. Aftershocks occurred within a zone 10 km long that included themain shock as well as on a parallel zone of seismicity about 5 km to the west that had beenintermittently active for the preceding year.

As is common in other volcanic and geothermal areas, many of the earthquakes in oradjacent to the Yellowstone caldera occur in swarms (Smith and Braile, 1994; Waite and Smith,2002; Farrell, 2007),. These swarms are characterized by numerous small, generally shallowearthquakes clustered in both space and time, without a mainshock, and having magnitudedifferences 0.25. Commonly these swarms are aligned on tectonic trends, though not necessarilyon recognized faults, and are generally considered to be related to migration of hydrothermal fluidsor magma in the shallow crust. A particularly intense earthquake swarm (fig. 14) occurred inOctober and November of 1985 a short distance northwest of the Yellowstone caldera (Waite andSmith, 2002). With time, the swarm activity migrated both away from the caldera and, to a lesserextent, downward as the caldera itself underwent a change from uplift to subsidence. The swarmwas likely to have been related to the subsurface movement of hydrothermal fluids or magma frombeneath the caldera, deflating the caldera area and possibly penetrating the shallow crust nearby asmagmatic dikes that did not breach the surface in eruption (Waite and Smith, 2002). Possiblemigration of magma may have induced local migration of hydrothermal fluids and magmatic gases(Husen and others, 2004).

Induced or triggered earthquakes constitute another class of seismicity observed recently inYellowstone (Husen and others, 2004). Such earthquakes are triggered by the passage of large-amplitude surface waves from distant sources to produce local transient dynamic stress changes asgreat as 0.15 MPa. A series of local earthquake swarms was observed, and clear changes in geyseractivity occurred immediately following local arrival of seismic waves from the 2002 Denali fault,Alaska, earthquake (M=7.9) in the Yellowstone National Park area, 3100 km from the epicenter.Beginning within hours of the arrival of surface waves, the YVO network located more than 250earthquakes in the first days after the Denali event. The eruption frequency of several geysers wasaltered, and numerous earthquake swarms close to major geyser basins were unusual in occurringsimultaneously in different geyser basins. This behavior is interpreted as having been induced bydynamic stresses associated with the passage of large-amplitude surface waves, perhaps locallyaltering the permeability of hydrothermal plumbing by unclogging existing fractures. Furthermore,redistribution of hydrothermal fluids and locally increased pore pressures triggered earthquakes in

11

these swarms. Although earthquake triggering and changes in geyser activity have beendocumented elsewhere, these notable changes at Yellowstone induced by a large-magnitude eventat such a great distance indicate the profound effects that remotely triggered seismicity can have onhydrothermal and earthquake hazards and also suggests that earthquakes and hydrothermalexplosions at Yellowstone might be triggered by local earthquakes in the future.

Crustal deformationMajor surface deformation has been recognized in Yellowstone since precise leveling

surveys were repeated in 1975-1977 after a lapse of 50 years. Caldera-wide uplift to a maximum ofat least 73 cm (fig. 15) was documented in these surveys (Pelton and Smith, 1979, 1982). Similaruplift continued, as measured repeatedly in several following years (fig. 16), averaging 23 mm/yfrom 1976 to 1983, then slowing down for about a year before subsiding as much as 35 mm/y until1987 (Dzurisin and Yamashita, 1987; Dzurisin and others, 1990).

Since these surveys, newer methods, including the Global Positioning System (GPS) andInterferometric Synthetic Aperture Radar (InSAR) have shown that surface deformation remainsactive over the entire area of the Yellowstone caldera and the seismically active area to thenorthwest, including the epicenter of the 1959 Hebgen Lake earthquake. Both uplift andsubsidence occur in this area, sometimes in a caldera-wide fashion and at other times in morecomplex patterns of more local uplift and subsidence (Meertens and Smith, 1991; Wicks andothers, 1998; Puskas and others, 2002; Wicks and others, 2006; Puskas and others, 2007).

Currently, deformation monitoring of the Yellowstone region is carried out by YVO mainlythrough a network of 16 GPS receivers (Yellowstone Volcano Observatory, 2006), of which 12 arelocated within Yellowstone National Park (fig. 17), continuously recorded at the University of Utahand by the EarthScope Plate Boundary Observatory. Additional stations are planned for futureinstallation. Each station records a time series of 2 horizontal components and 1 verticalcomponent sampled every 5 seconds, which are analyzed at the University of Utah to producecoordinate solutions. In addition, about 60 GPS stations are recorded in special surveys rather thanbeing monitored continuously. Repeat leveling surveys also continue intermittently.

Together, the geodetic data reveal complex patterns of crustal deformation over a period ofdecades (fig. 14), as summarized by Puskas and others (2007). From 1923 to 1985 caldera-wideuplift occurred at an average rate of 22 mm/yr, changing in 1986 to subsidence. Subsidencecontinued within the Yellowstone caldera to 1995 at 14 mm/yr while uplift occurred northwest ofthe caldera at 5 mm/yr. In 1995, caldera uplift resumed at 9 mm/yr, changing again in 2000 tosubsidence at 9 mm/yr while uplift continued northwest of the caldera at an average of 12 mm/yr.A sudden change in 2004 to caldera uplift proceeded at unprecedented rates of up to 60 mm/yr,continuing into 2007. These rapid changes in motion across the Yellowstone caldera over morethan a half century clearly reflect the importance of transient large-scale crustal deformationsources that probably include both the subcaldera magma chamber and localized zones ofpressurized hydrothermal fluids that inflate and deflate the surface as the locations and properties ofthe sources change with time.

Detailed analysis of ancient shorelines of Yellowstone Lake, both subaerial andsublacustrine, indicates that similar changes of major uplift and subsidence date back at least as faras the final retreat of Pleistocene glaciers from the lake basin (Meyer and Locke, 1986; Locke andMeyer, 1994; Pierce and others, 2002; Johnson and others, 2003). Pierce and colleagues (2002)conclude that cycles of uplift and subsidence have occurred repeatedly in the Yellowstone Lakebasin during the past 14,000 years; the record of these cycles is superposed on the record of a long-term decrease in lake level by downcutting of the lake outlet, but that little net elevation changeoccurred between about 14 ka and 3 ka (fig. 18).

12

Together, earthquake focal mechanisms and both vertical and horizontal deformationpatterns suggest that the crustal stress field recorded both by earthquakes and by surfacedeformation are generally consistent with regional basin-range extension. In the vicinity of theYellowstone caldera, however, the stress field is strongly modified by the effects of migratingmagmatic and hydrothermal fluids to produce changes in seismicity, uplift, and subsidence on timescales of months to a few years (Dzurisin and others, 1990; Dzurisin and others, 1994; Puskas andothers, 2002; Waite and Smith, 2002; Husen and others, 2004; Waite and Smith, 2004; Wicks andothers, 2006; Puskas and others, 2007).

Hydrothermal and gas activity

At each eruption, immediately preceding, was an upheaval of some fifty feet high, followed byone great explosion in which the water was thrown two hundred fifty to three hundred feet andfrequently hurling stone one foot in diameter five hundred feet from the crater.

Joaquin Miller (in Muir, 1888), describing hydrothermal explosions of Excelsior Geyser.

Yellowstone National Park hosts one of the largest hydrothermal systems on Earth and, inparticular, has more geysers than all others in the world together. Most of these hot springs,geysers, and fumaroles present hazards only to visitors who may stray from established pathwaysinto areas of unstable ground and hot water; it is impractical to consider such daily hazards in thecontext of the present assessment. One of the most common acute geological hazards atYellowstone, however, consists of shallow-rooted explosions of steam, water, and rock withoutany associated volcanism. These hydrothermal explosions (or eruptions) occur when hotsubsurface waters flash to steam, violently breaking the confining rocks and ejecting them from anewly formed or existing crater. Individual events can last as briefly as a few seconds or as long asseveral hours, and intermittent explosive activity can continue for years.

Hydrothermal-explosion craters are found in Yellowstone National Park at scales rangingfrom less than a meter to several kilometers in diameter (Muffler and others, 1971; Morgan andothers, in review). Visitors to popular thermal areas like Upper and Lower Geyser Basins, NorrisGeyser Basin, and West Thumb Geyser Basin see circular, funnel-shaped pools, often with jaggedwalls, filled with thermal water (fig. 19). Many of these pools appear to have formed inhydrothermal explosions (Marler and White, 1975). Porkchop Geyser in Norris Geyser Basin,Excelsior Geyser in Midway Geyser Basin, and Seismic Geyser in Upper Geyser Basin haveproduced well-documented historical explosions (Marler and White, 1975; Fournier, 1989;Whittlesey, 1990). Geologic research has identified many large craters formed before the time ofhistorical records, including a number at and near the north edge of Yellowstone Lake (fig. 20:Mary Bay, Indian Pond, and Turbid Lake; also see fig. 24). If similar large explosions were tooccur today in that vicinity, they could threaten infrastructure and visitor safety along the EastEntrance Road, at Fishing Bridge and as far away as Lake Village. Geologic and historic evidencein Yellowstone and elsewhere suggests that such hazardous activity could last days, weeks or evenyears.

Browne and Lawless (2001) provide a lengthy summary of current knowledge ofhydrothermal “eruptions” around the world, preferring the word eruption over explosion toreinforce the idea that the events were part of a continuum extending from geysering (no rocksejected) to very large rock-hurling eruptions creating craters hundreds to thousands of metersacross. While acknowledging the terminology of Browne and Lawless (2001) but to avoidconfusion with magmatic eruptions, we continue to use the term hydrothermal “explosions” in thishazards assessment for events that can form craters or eject rocks.

We follow Browne and Lawless (2001) in distinguishing hydrothermal eruptions/explosionsfrom other types of ground water eruptions such as:

13

Phreatic eruption: An eruption of steam, water and rock caused when cool ground water isheated rapidly to its boiling point by magmatic heat but no magma is actually erupted.

Phreatomagmatic eruption: Similar to a phreatic eruption but with evidence for eruption ofmagma as well as fragments of the confining rock.

Magmatic-hydrothermal eruption: An eruption caused by magma heating a pre-existinghydrothermal system; the magma may or may not reach the surface as a part of the ejectedmaterials. An excellent example occurred in New Zealand in 1886, when intrusion of basalticmagma into a long fracture system caused hydrothermal explosions as far as 20 km away from thelava vent (Hedenquist and Henley, 1985; Simmons and others, 1993).

In contrast, hydrothermal eruptions/explosions occur in pre-existing hydrothermal systemswithout the proximate influence of magma. It appears that all explosions at Yellowstone since thelatest glaciation, roughly the past 16,000 years, are true hydrothermal explosions with no directparticipation of magma.

Pressures, temperatures, and fluids in geothermal systems

Understanding the causes of hydrothermal explosions requires understanding the structureand dynamics of geothermal systems. Such systems are regions of anomalous heat flow in theshallow crust of the Earth. Geothermal systems can be divided into two distinct types: hot-water(or liquid-dominated) systems and vapor-dominated systems (White and others, 1971).

In a liquid-dominated system, the Earth’s subsurface is saturated with hot water, which islocated in interconnected pore spaces and fractures. Liquid-dominated systems can have two typesof surface manifestations: a) hot springs, which can represent the direct outflow of hot water fromthe geothermal reservoir, and b) fumaroles, which represent steam and gas boiled off the reservoir.At Yellowstone National Park, the hot springs of liquid-dominated systems are typically neutral orslightly alkaline, Cl-rich, and are saturated with silica (SiO2). As these waters cool, they precipitateamorphous silica as deposits called siliceous sinter. Waters that precipitate silica are common in,but are not limited to, the Upper and Lower Geyser Basins, the Norris Geyser Basin, the WestThumb Geyser Basin (Fournier, 1989) and sublacustrine hydrothermal vents in Yellowstone Lake(Shanks and others, 2005; Shanks and others, in review).

Vapor-dominated systems lack sufficient water for complete liquid saturation of thereservoir, resulting in low-density steam and gas being the interconnected phase. Vapor-dominatedreservoirs can still contain an appreciable amount of liquid, which remains within pore spaces inthe host rock. Surface manifestations above vapor-dominated reservoirs typically consist of “acid-sulfate” features such as fumaroles, acid pools, and mud pots. Hot springs issuing neutral watersare usually absent. Instead, acid pools and mud pots form due to condensation and oxidation ofacid gases and steam in perched bodies of surface and near-surface water. Most of the largethermal areas in the eastern part of Yellowstone National Park (e.g., Mud Volcano, Hot SpringsBasin, Sulphur Hills, Josephs Coat) formed above vapor-dominated geothermal areas and haveacid-sulfate soils and widespread alteration of surface rocks, commonly to clays (White and others,1971; White and others, 1975).

The simplest geothermal reservoir at Yellowstone consists of ground water at its boilingtemperature for a given hydrostatic head (fig. 21). In these situations, pressure is dictated by theweight of the interconnected column of subsurface and surface water. Under normal conditions,the confining pressure is sufficient to prevent the water from boiling catastrophically. Geysereruptions occur when subtle boiling at the top of the column induces local depressurization andconsequent boiling further down the column. Conditions such as a landslide, earthquake, or dam-break may cause a large and sudden decrease in pressure; geothermal water then may becomehighly superheated relative to its boiling temperature at the new, lower pressure, causing rapidconversion of some of the water to steam, forcing expansion, and triggering an explosion.

14

Hydrothermal explosions can be initiated more readily when pressures exceed hydrostatic. Ideally,because the density of hot water is less than that of cold water, hydrothermal systems should beslightly under-pressured with respect to the cooler surrounding ground water systems. However,mineral-precipitation reactions, especially those forming silica, serve to clog permeable pathwaysin the aquifer and can cause the geothermal waters to become isolated with respect to theirsurroundings. Moreover, the presence of gas and steam can increase the pressure of the system. Asa result, many geothermal drill holes, including research holes drilled at Yellowstone in the 1960s(White and others, 1975), are positively pressured with respect to the predicted hydrostatic gradient(fig. 22).

An additional factor increasing the pressure within geothermal areas is the buildup of gasessuch as carbon dioxide (CO2) and hydrogen sulfide (H2S), which do not condense upon ascent andcooling and can build up in concentration beneath an impermeable caprock (Hedenquist andHenley, 1985). At the vapor-dominated Ngawha system in New Zealand, for example,extrapolations of deep gas pressures to near surface conditions could allow a hydrothermal eruptionthat would lift and disperse about 130 m of overlying rock (Browne and Lawless, 2001). Thus,vapor-dominated reservoirs also can experience hydrothermal explosions, either owing to buildupof non-condensable gases or simply by release of superheated steam. In general, explosions fromvapor-dominated reservoirs are thought to be somewhat less dangerous than liquid-dominatedreservoirs because a given volume of steam will contain less potential energy than the same volumeof liquid at the same temperature. Even in vapor-dominated reservoirs, most of the potentialenergy of the system resides within the residual liquid water (Browne and Lawless, 2001).

Mechanisms of hydrothermal explosion

The mechanisms for hydrothermal explosions recognized by Browne and Lawless (2001)are summarized here in the context of the geothermal system at Yellowstone.

Pressures exceeding lithostatic: A hydrothermal explosion can occur if fluid pressure isregulated by an impermeable caprock, such that pressures increase until they exceed the weight andstrength of the overlying rock. This mechanism is thought to be comparatively rare at geothermalfields, where measured pressures are usually well below lithostatic, but eruptions from these fieldsare relatively common. Moreover, the common hot springs and fumaroles at the surface abovegeothermal fields imply that their caprock is not generally impermeable. At Yellowstone, it is clearthat areas affected by hydrothermal explosions are those having already-established reservoirsconnected to distinct surface expressions. Browne and Lawless (2001) infer that this mechanism ismost important above young geothermal reservoirs as they start to interact with overlying, as yetunaltered, surface rocks. Phreatic eruptions at reactivating volcanoes also might occur whentransients in ground water pressures exceed the lithostatic load. Potentially, self-sealing due toprecipitation of minerals from hydrothermal fluids could decrease the permeability of hydrothermalaquifers, causing pressures to approach lithostatic.

Slow accumulation of steam and/or gas: Pressures can more regularly exceed lithostatic ifsteam is present in the system and can ascend, transmitting pressure to shallower regions. This isthought to be the common mechanism for explosions in exploited geothermal fields (Browne andLawless, 2001) when withdrawal of geothermal fluid lowers subsurface pressures, therebytriggering additional boiling. Because cooling of noncondensable gas has a negligible effect on itsvolume, its presence will increase the likelihood of overpressure. Long-term drought and a drop inground water levels could cause a similar phenomenon wherein shallow steam pressures increasewith a decrease in the elevation of the ground water table (fig. 23, adapted from fig. 6 of Browneand Lawless, 2001).

Rapid subsurface pressure release: Any geothermal system that follows the boiling pointwith depth can be triggered into an eruption by a sudden release in pressure that causes flashing of

15

liquid water to steam. Such a pressure release could result from an earthquake, a landslide,draining of a lake, deglaciation, or a hydrothermal fracturing event within the reservoir. AtYellowstone, such events are known to have caused hydrothermal perturbations. The 1959 HebgenLake earthquake (M=7.5) induced eruptions of some 289 springs, including 150 with no previousrecord of geysering (Marler, 1973). Most of these eruptions were solely of liquid water and steam,though Marler and White (1975) document the growth of a new feature, Seismic Geyser, whosegenesis involved a series of rock-hurling episodes. Muffler and others (1971) postulated that thevery late Pleistocene Pocket Basin explosion in the Lower Geyser Basin was initiated by drainageof a postglacial lake, causing a pressure drop in the underlying geothermal reservoir.

Addition of magmatic heat or gas: Addition of external magmatic heat or gas yields what istermed a magma-hydrothermal eruption. One might consider all heat at Yellowstone to beultimately magma-derived, but the term is used here to denote a rapid transfer of heat from ashallow magmatic intrusion directly into the geothermal system. Though such a mechanism isunlikely for the small shallow explosions most common at Yellowstone, its importance cannot beruled out for some of the earlier large hydrothermal explosions of the late Pleistocene and earlyHolocene. If magma, however, did induce some of the large explosions, it did so without reachingthe surface.

Progressive flashing: Browne and Lawless (2001) conclude that this is the most commontype of hydrothermal explosion. It initiates close to the ground surface and works its waydownward with time as rocks fracture, causing the boiling front to move down, resulting inincreased boiling, brecciation, and depressurization of the underlying system. The model requiresthat boiling water exists near the surface and overlies water at a boiling-point-for-depth temperaturegradient. As such, the initial confining pressure can be minimal, and open pools can be the site ofinitiation. The eruption itself might be triggered by events like those described in the subsurfacepressure release model, but with the boiling water column destabilizing much closer to the surface.

Hydrothermal explosions in Yellowstone

At Yellowstone, a geologic record of hydrothermal explosions exists only for eventsoccurring after the most recent glaciation (ending about 16,000 years ago, Pierce and others, 2002),which effectively erased evidence of any earlier such events. Evidence for eighteen large (>100-m)hydrothermal explosions (table 3; fig. 24) is found within Yellowstone National Park (Morgan andothers, in review). Many large explosion craters are present on land, but others have beenidentified in Yellowstone Lake (Wold and others, 1977; Morgan and others, 2003b; Morgan andothers, in press-b; Morgan and others, in review). Most of the largest explosion craters, such asMary Bay, Turbid Lake, Duck Lake, Indian Pond, and Pocket Basin (fig. 24), are found within andalong the margin of the 640-ka Yellowstone caldera. A few are present along the tectonicallycontrolled, north-trending Norris-Mammoth corridor. The hydrothermal-explosion craters appearto be relatively shallow features affecting only the upper several hundred meters of substrata, whichtypically have been previously affected by hydrothermal alteration (Muffler and others, 1971;Morgan and others, in review). In Yellowstone, no large hydrothermal explosion is associated witha volcanic event, and no evidence exists for a hydrothermal explosion triggering any volcanism.

Figure 25 displays historical hydrothermal-explosion sites identified through a literaturesearch. As with the prehistoric craters and deposits, the historical eruption sites are concentratedwithin the caldera and the Norris-Mammoth corridor. Notably, the Upper and Lower GeyserBasins hosted many of the observed events. Appendices 1 and 2 provide descriptions of some ofthe historically observed explosions and the prehistoric explosion deposits studied.

The end result of most hydrothermal explosions is a crater, commonly water-filled,surrounded by a berm of fragmental rocks with steep inward slopes and gentler outer slopes. Thedeposits generally comprise pieces of hydrothermally altered or mineralized materials such as

16

siliceous sinter, mud, breccia, and subsurface lithologies such as lake, beach, or glacial sedimentaryrocks and rhyolitic rocks from deeper levels.

The individual fragmental deposits are quite varied and reflect the stratigraphy ofunderlying rocks and the depths of rocks evacuated by the explosion, and they commonly recordcomplex, multiple-event histories. Silica, in the forms of quartz, chalcedony, opal, and amorphoussilica, is the most common cement of the explosion breccias and cross-cutting veins, but pyrite,calcite, and zeolites also are common hydrothermal minerals.

The complexity of the deposits implies repeated explosion events over a wide range ofscales, with lithic fragments being ejected, falling back into or adjacent to the vent, and beingcemented with silica to form breccias. Multigenerational heterolithic breccias also are common,indicating an extended process of repeated brecciation and hydrothermal cementation that likelyoccurs in hydrothermal systems well below the explosion vents (Keith and Muffler, 1978; Morganand others, in review).

Factors contributing to hydrothermal explosions

Several factors contribute to the likelihood of hydrothermal explosions. Prior to theeruption of Porkchop Geyser in 1989, Fournier and others (1991) had interpreted chemicalindicators as showing that the temperature of the water feeding the hot-spring pool had beenincreasing with time. Increased temperature of the deep water would have had two primary effects.First, it would have increased the amount of boiling and therefore steam production as the waterrose towards the surface. Second, it would have increased the amount of silica in the ascendingwater and thus the supersaturation of silica as the water cooled during its rise. The first factorwould have increased pressure in the system as the steam/liquid ratio in the subsurface increased inthe relatively constant-volume system. The second factor would have decreased permeability in thegeyser’s conduit, potentially causing a decrease in the rate at which water could flow through thesystem. Thus, increased heat to the system increased the likelihood of pressurization andhydrothermal explosions.

Another influence on the likelihood of hydrothermal explosions is the depth of the vapor-liquid interface above a liquid-dominated geothermal reservoir (fig. 23), as this depth controls thepressure of any vapor reservoir near the surface. If rocks of low permeability overlie a vapor-dominated region, then the pressure can rise if any or several of the following things shouldchange: a) increased heat supplied to the system, b) increased amount of gas or steam risingthrough the system, or c) reduced ground water recharge to the system causing the geothermalwater table to fall. Any of these factors would cause an increase in the thickness of the steamreservoir, resulting in greater pressures transferred toward the surface through the vapor/steamreservoir. This process occurs commonly at geothermal wells, where gas must be “bled off” toprevent displacement of water and lowering of the water level within the well. If gas or steam isallowed to accumulate in idle geothermal wells, the results can be disastrous. Because the groundsurface at Yellowstone is generally permeable above vapor-dominated areas, allowing abundant gasflux (Werner and Brantley, 2003), this mechanism may not be a leading cause of explosions atYellowstone but might be so on occasion.

Rapid pressure reduction is commonly invoked as a cause of hydrothermal destabilization.Withdrawal of fluids from geothermal wells is documented to have lowered subsurface pressuresand induced boiling and hydrothermal explosions in geothermal fields (Scott and Cody, 2000).Earthquakes, deglaciation, or lake drainage all could cause sufficient depressurization of a boilingaquifer to cause hydrothermal explosions at Yellowstone (Muffler and others, 1971; Morgan andothers, in review). Bargar and Fournier (1988) demonstrated that parts of the geothermal reservoirwere superheated by 20-50°C following deglaciation, at about 12 to 15 ka. This is presumablybecause while glacial ice was present, areas beneath the ice-rock contact were pressurized

17

compared with ice-free conditions, allowing a higher boiling temperature for H2O. Afterdeglaciation, considerable boiling would be necessary to re-equilibrate the water column to thenew, lower-pressure conditions. Although this temporary instability would not likely be a directtrigger for hydrothermal explosions, it could be a possible contributing factor; whether anyexplosions resulted directly from deglaciation and attendant depressurization is not known.

Earthquakes, extremely common at Yellowstone, are also known to have strong effects ongeothermal features (Watson, 1961; Marler, 1964; Marler and White, 1975; Husen and Smith,2004). Some of the large prehistoric hydrothermal explosions could have been triggered bydestabilization due to passage of large-amplitude seismic waves that dynamically increase localstress in hydrothermal reservoirs.

Although several factors can be identified as potential triggers for hydrothermal explosionsat Yellowstone, the reality is that there are only sparse observational or scientific monitoring dataon past explosive events. Geologic evidence for triggering mechanisms for past events is rare andambiguous though current studies are aimed at detecting such information where it may exist.

Toxic gases

After steam, carbon dioxide (CO2) is the most common constituent of volcanic gas and canbe emitted in sufficient quantities to pose a hazard at many volcanic and geothermal systemsaround the world (Baxter, 2005). After subtracting steam, CO2 typically constitutes 95 to 98% ofthe gas emitted from Yellowstone’s fumaroles and bubbling pools (Werner and Brantley, 2003).Carbon dioxide is non-toxic in low concentrations and makes up about 0.038% of the Earth’satmosphere. However, because it is about 50% heavier than normal air, it can accumulate to muchhigher concentrations in soils and low or protected areas such as valleys and caves. Carbon dioxideconcentrations of >10% are toxic to humans and animals. When air contains over 20-30% CO2,even a few breaths can quickly lead to unconsciousness and death from acute hypoxia, severeacidosis, and respiratory paralysis (Hill, 2000). Hypoxia is a condition in the body resulting fromthe displacement of oxygen such that it inhibits normal metabolism. Acidosis occurs when CO2

acidifies the blood, causing irreversible cellular damage. In the Yellowstone region, the localconcentration of CO2 to potentially toxic levels is generally only temporary and is restricted toconfined or topographically low areas of hydrothermal activity, generally under windlessconditions.

Volcanoes commonly emit acid gases like sulfur dioxide (SO2) and hydrogen chloride(HCl). For example, at the volcanoes Kilauea in Hawaii and Masaya in Nicaragua, these acid gasesform aerosols that are a chronic hazard to both local vegetation and human populations (Baxter,2005). At Yellowstone, the hydrothermal system and its host rocks act as a buffer that absorbs andneutralizes acid gases, forming hydrogen sulfide (H2S) gas and soluble sodium chloride.Additional H2S may rise directly off the magma. The concentration of H2S is typically about 20 to200 times less than that of CO2, but its toxicity is much greater. Though H2S has an extremelystrong “rotten-egg” odor at levels of only a few parts per billion (ppb), concentrations of more than10 ppm in the air will rapidly deaden the human sense of smell to its presence (Mandavi, 2005).Concentrations of 100 ppm can cause severe eye and throat irritation, and at concentrationsexceeding 700 ppm loss of consciousness and death can occur rapidly. Hydrogen sulfide forms acomplex bond to iron in mitochondrial cytochromes, thereby arresting aerobic metabolism in aneffect similar to cyanide toxicity (Milby and Baselt, 1999).

The HazardsThe geologic setting, geophysical activity, and hydrothermal systems reviewed above

provide the framework within which to consider potential hazards arising from any future volcanic,

18

hydrothermal, or gas-emission activity. For each of these categories of potential hazards,discussion is organized in terms of different types of activity that might develop as the Yellowstonemagmatic-tectonic-hydrothermal system continues to evolve.

Volcanic-eruption hazardsAs is characteristic of many large continental magmatic systems, eruptive activity in the

Yellowstone Plateau volcanic field is highly episodic and involves long periods of time betweeneruptions; many of the eruptions are quite large. Such systems are particularly difficult to evaluatein terms of the probabilities of hazardous events and the consequent risks to people and resources.In order to bring some coherence to this discussion, it is organized by considering first the smaller,more likely future volcanic eruptions and proceeding to the larger and potentially most destructivebut least likely events.

Basaltic eruptionsAs noted earlier, basaltic lavas have erupted around the margins of the active, mainly

rhyolitic Yellowstone Plateau volcanic field throughout its evolution. The absence of basalts fromwithin the rhyolitic source areas is interpreted to reflect the trapping within the crust of any basalticmagmas that might have intruded from zones of partial melting in the upper mantle beneath crustalrhyolitic magmas of lower density (Christiansen, 2001). Only after about a million years havebasalts erupted through the cooled, crystallized, and fractured upper-crustal magmatic sources ofthe first and second rhyolitic cycles; no basaltic vents, however, occur within the third-cycleYellowstone caldera. A few small outcrops of basalt do occur on the northwest caldera wall nearPurple Mountain (Christiansen and Blank, 1974), but they are erosional remnants of lavas thatflowed down the steep slope from vents farther north. Additionally, some rare quenched inclusionsof basaltic magma were found within the basal part of the rhyolitic West Yellowstone flow near thecrest of the Madison Plateau west of Little Firehole Meadows (R. L. Christiansen and H. R. Blank,Jr., unpubl. data), suggesting that basaltic magmas from lower-crustal levels might have played arole in mobilizing some intracaldera rhyolitic magmas for eruption.

Most postcaldera basalts of the Yellowstone area are glaciated erosional remnants of once-larger flows (fig. 10). Extrapolation to the likely outlines of initial distributions suggests that mostindividual basaltic eruptions covered areas of less than about 5 km2 and did not exceed 0.1 km3 involume. The largest single basaltic flow field, however, the Falls River Basalt, extends from thesouthwestern caldera margin to Henrys Fork of the Snake River, a distance of about 60 km. TheFalls River Basalt flow field may have covered 900 km2 and may account for an eruptive volume ofnearly 20 km3. At least two vents and possibly more fed the flow field, only one of which is nowexposed (fig. 10). The next largest basaltic flow field near Yellowstone National Park, the GerritBasalt covering about 100 km2 in the area of Island Park, west of Yellowstone, erupted from atleast 13 vents. Basalts cover the Eastern Snake River Plain west of Island Park, and at least 4 ventsfor basalts of the Snake River Group are within or immediately adjacent to the basin of Island Park(Christiansen, 1982).

The postcaldera basalts of the Yellowstone region—generally pahoehoe flows—are mainlyolivine tholeiites having a range of K2O contents (but commonly <0.5%). Lavas producing suchflows would be expected to have low viscosities and to erupt rapidly. Probably most of the basaltserupted in events lasting no more than a few weeks to a few months, but a large flow field like theFalls River Basalt might have accumulated in multiple eruptions, each lasting many months. Mostof the recognized basaltic vents are localized agglutinated scoria accumulations or small lavashields, but some basalts may have originated as dike-fed fissures. A few formed cinder cones.

19

A total of 33 postcaldera basaltic vents (fig. 10) have been recognized or inferred in the areaimmediately surrounding the Yellowstone caldera (Christiansen, 2001; Smith and Bennett, 2006).Of these, 17 are within or adjacent to the Island Park basin; the others are scattered around allsectors of the Yellowstone-caldera margin. Because of their generally scattered distribution, mostof the vents probably remain preserved at the surface, but some might have been buried by youngermaterials or not recognized during geologic mapping. If it is assumed that 80 percent of the actualvents have been recognized, there could have been as many as 40 basaltic eruptions around theperiphery of the Yellowstone caldera in the past 640,000 years. Of these about half occurred inIsland Park. No postglacial basaltic eruptions have been recognized, indicating that none hasoccurred within the past 16,000 years.

On the basis of the foregoing, the average period between basaltic eruptions in the areaaround Yellowstone National Park during post-Lava Creek time is about 16,000 years. Theaverage annual probability (i.e., the number of events divided by the time period, in years, duringwhich they have occurred) of a basaltic eruption occurring somewhere around the periphery of theYellowstone Plateau volcanic field is therefore 6x10-5. However, it is unclear whether the basalticeruptions, like some of the Yellowstone rhyolitic eruptions, may have been clustered in time; if so,the long-term average recurrence period may have little direct bearing on future eruptionprobabilities. The most likely location for any such future basaltic eruption is within the basin ofIsland Park but could be anywhere else within a band about 40 km wide surrounding theYellowstone caldera. Any such eruption is most likely to last between a few weeks and severalmonths. It is possible but unlikely that basalt could erupt from within the caldera; if such an eventwere to occur, however, it would signal the demise of the large Yellowstone-caldera rhyoliticmagma chamber.

The principal hazard likely to result from a basaltic eruption around the periphery of theYellowstone caldera would be coverage of an area of several square kilometers by lava, one to afew tens of meters thick. In addition, basaltic ash and cinders from the eruptive vent might blanketareas of many hundreds of square kilometers to depths of a few meters to a few centimeters,decreasing in thickness outward from the vent in directions determined by the prevailing winds. Ifa basaltic vent were to emerge from beneath shallow water or a large area of saturated ground,phreatomagmatic eruptions could produce pyroclastic surges within the proximal area that couldblast down trees and cause other similar destruction.

In addition to any primary hazards of lava inundation and ash blanketing, there are likely tobe secondary hazards from any basaltic eruption in the Yellowstone region. In particular, such anevent would be likely to start fires around an advancing flow front, particularly under dryconditions. Debris flows or floods could be triggered by the melting of any significant snow packor by temporary blockages of major drainages and subsequent release of floodwaters as theblockage was undermined or overtopped by rising waters.

Given the ongoing YVO monitoring program, it is likely that there would be recognizablepremonitory indicators of any impending basaltic eruption. In particular, multiple shallowearthquake swarms focused in a small, probably linear area would be likely to be followed oraccompanied by volcanic tremor, as has been observed at many basaltic volcanoes as they prepareto erupt. It is likely that surface ground cracks would open in the immediate stages prior to anybasaltic eruption as a dike approached the surface. Because YVO deformation monitoring isfocused on the Yellowstone caldera, initial uplift associated with the shallow intrusion of basalticmagma peripheral to the caldera might not be recognized quickly.

The emission of magmatic gases to the surface would be a major indicator of impendingeruptive activity but might be quite difficult to recognize in the presence of Yellowstone’s massivehydrothermal system, which tends to scrub out magmatic gases passing through it (Symonds andothers, 2001). Nevertheless, any locally increased emissions of CO2 or H2S should be monitored

20

closely; any appearance of SO2, presently absent at Yellowstone, would be especially indicativethat the hydrothermal system was becoming dried out by shallow magmatic intrusion. Evenwithout SO2, significant localized increases in the ratio of sulfur gases to carbon gases wouldsuggest the possibility of magmatic gas reaching shallow subsurface levels.

Because no basaltic eruptions have occurred in more than 16,000 years at Yellowstone,there are no well-established magmatic pathways. Thus, it is most likely that premonitoryseismicity would be sufficiently prominent and of long enough duration to allow temporarymonitoring of ground deformation and gas emission to provide additional information forinterpreting possible locations and the nature of any such eruptive event. Among the earliestindicators might be relatively deep long-period seismicity. The most immediate precursorsprobably would occur as a shallow intrusion entered the hydrothermal system, generating veryactive short-period seismicity and possibly hydrothermal explosions. However, the distinctionbetween an impending basaltic or small rhyolitic eruption might be difficult to evaluate before theinitial venting.

Because the basaltic lava flows of Yellowstone are virtually all tube-fed pahoehoe, theiradvance would probably be slow enough to allow mitigating measures to be taken except, perhaps,in areas close to erupting vents.

Large rhyolitic eruptionsAt least 17 large rhyolitic lava flows, most of them with volumes of 10 km3 or greater, have

erupted within the Yellowstone caldera during about the past 170,000 years (Christiansen, 2001).Stratigraphically they belong to the Central Plateau Member of the Plateau Rhyolite. Each of theselava flows extruded through one of two linear vent zones that cross the caldera along theextrapolated positions of extracaldera tectonic fault zones (figs. 8, 11), and activity has beenessentially contemporaneous on both zones. The largest of these lava flows cover areas greaterthan 350 km2 and have volumes greater than 30 km3 (table 2). The constructional topographyformed by these flows is represented by the Pitchstone, Madison, and Central Plateaus (figs. 1 and26). The Pitchstone-Madison Plateau alignment lies on the extrapolated position of the Tetonnormal-fault zone (fig. 11) and extends to the southwest edge of the tectonic West Yellowstonebasin (fig. 8). The Central Plateau alignment (fig. 8) lies on an extrapolation of both the Sheridannormal-fault zone (fig. 11) and the extracaldera Norris-Mammoth corridor (fig. 10).

As these young intracaldera rhyolite flows are petrographically, chemically, and isotopicallydistinct from early intracaldera rhyolites (Hildreth and others, 1984; Hildreth and others, 1991) andhave an age range separated by about 80,000 years from the youngest known earlier intracalderaflows, they are considered here together as a group separate from the older postcaldera lavas. Thehazard potential for a possible future intracaldera eruption may be best reflected in thecharacteristics of this group of voluminous rhyolites of ~170 ka and younger.

Available K-Ar and 40Ar/39Ar dating of rhyolitic flows of the Central Plateau Member (table2; figs. 9, 26) indicates that they erupted in a few major episodes. The earliest of these episodesprobably began with the Mallard Lake flow that erupted from the Central Plateau vent alignmentand covered the eastern part of the Mallard Lake resurgent dome (Christiansen, 2001). It was eitheraccompanied by or immediately followed by renewed uplift of the Mallard Lake dome.Subsequent eruptions, within a period of less than 10,000 years, produced the Dry Creek, WestThumb, and Mary Lake flows along the Central Plateau vent zone and the Buffalo Lake flow alongthe Madison Plateau vent zone (fig. 9). Additional rhyolitic lavas, now buried, may well haveerupted during this time from either or both of those zones. In addition to these lavas, at least onemajor pyroclastic eruption occurred along the Central Plateau vent alignment to produce the tuff ofBluff Point. Although its volume cannot be reconstructed convincingly because of erosion andburial by younger lavas, eruption of this pumiceous pyroclastic unit was sufficiently large to

21

produce a source caldera 10 km in diameter, now represented by West Thumb, the westernmostbasin of Yellowstone Lake (fig. 8). The eruptive volume almost certainly was several tens of km3.The actual time span represented by eruptions of these oldest units of the Central Plateau Memberis not defined precisely. The nominal span of weighted-mean ages for the dated lava flows and tuffof this group is 173±11 to 160±3 ka, but disagreements between individual age determinations andthe stratigraphic order interpreted from geologic mapping suggest that some of the isotopic ages areincorrect. The group mean age and standard deviation of the individual weighted-mean isotopicage determinations on all of these units is 167±5 thousand years, and all the weighted-mean ages ofthe individual units of this group overlap that range. Thus, the range might be a closerrepresentation of the actual span of time involved in erupting them than the nominal span ofindividual age determinations.

Distinct in time but only shortly after the first group of eruptions along the Central Plateauvent alignment (fig. 8) were the Aster Creek, Elephant Back, Spruce Creek, and Nez Perce Creeklava flows (fig. 9). The undated Spring Creek flow may have vented on the Madison Plateau ventalignment during this same episode. At least one significant pyroclastic unit, the tuff of ColdMountain Creek, is also interstratified among lavas erupted along the Madison Plateau alignmentalthough its source location is not known. This tuff may represent early pyroclastic activity fromthe vent for a large lava flow of the Madison Plateau. The age of this group of rhyolitic eruptions isabout 150±5 ka, calculated similarly to that of the earlier group.

Another relatively brief episode of large-volume rhyolitic eruptions within the MadisonPlateau vent zone formed the Summit Lake, Bechler River, and West Yellowstone flows (fig. 9);the smaller Douglas knob and Trischman Knob domes also probably erupted during this episode,possibly as late-stage vent domes of the Bechler River flow (Christiansen, 2001). The area coveredby flows of this age group is smaller than for the older groups, and all flows erupted during thissequence probably are represented by surface exposures. Nominal weighted-mean K-Ar ages ofthese lavas range from 124±10 to 114±1 ka, but just as for the older groups, stratigraphic relationssuggest that the actual time span of eruption was short. A mean age and standard deviation of theindividual weighted-mean age determinations on these flows is 118±5 ka.

Yet another episode of relatively large rhyolitic lava eruptions vented along the CentralPlateau at about 102-103 ka, producing the Hayden Valley and Solfatara Plateau flows (fig. 9).The age of these flows, calculated similarly to those of the older groups, is 103±1 ka.

The most recent episode of large intracaldera rhyolitic lava eruptions is represented by theadjacent Grants Pass and Pitchstone Plateau flows (fig. 9), with weighted-mean ages of 72±3 and79±11 ka, respectively. Both flows may well represent a single eruptive event, the linear vent forthe Grants Pass flow representing an early dike phase, and the Pitchstone Plateau flow issuing froma longer-lived central-vent phase. The weighted-mean age and standard deviation of these flows is76±5 ka.

Several factors suggest a model for systematic evolution for rhyolitic eruptions of theCentral Plateau Member that may have implications for related volcanic hazards. Volumes of theCentral Plateau lavas (table 2) were calculated from geologic maps using a cut-and-fill estimator inArcInfo, a Geographic Information System; those calculations proceed by estimating the originalextent and a computer-generated extrapolation of a flat bottom to each flow, resulting in aminimum-volume estimate and, thus, probably an underestimate of the total volume of all theflows. The aggregate volume of the oldest group, at 167±5 ka, is by far the greatest, at least 138km3. That volume, however, is possibly even greater, ~400 km3 as suggested by a separatecalculation of the total volume for all Central Plateau Member lavas of >600 km3, with thedifference between these calculations listed as “unobserved units” in table 2. (An earlier estimateof the total volume of the Central Plateau Member by Christiansen (2001) was 900 km3). Theaggregate volume of the second group of lavas, at 150±5 ka is at least 62 km3, and that of the next

22

group, at 118±5 ka, is about 86 km3. The volume of the 103±1-ka pair is about 9 km3, and that ofthe youngest pair, at about 76±5 ka, is about 70 km3. (These volume estimates do not include thetuffs of Bluff Point and Cold Mountain Creek, for which relevant data are inadequate). Thus, largevolumes of rhyolite erupted in episodes separated by roughly 15,000 to 30,000 years from each ofthe two linear extensional zones across the Yellowstone caldera where they intersect the calderaring-fracture system. More than 70,000 years have ensued since the latest eruptions.

Isotopic analyses of oxygen on most stratigraphic units of the Yellowstone Plateau volcanicfield were reported by Hildreth and others (1984). Oxygen, being by far the most abundantchemical element in any igneous rock, is particularly useful in tracing processes that affect themagmatic system as a whole. Although quartz phenocrysts in the oldest rocks of the volcanic fieldhave 18O >7‰—more or less typical igneous values—quartz in lavas erupted shortly after thefirst-cycle caldera-forming eruption had sharply depleted 18O, to values as low as ~4‰, resultingfrom massive assimilation of low-18O hydrothermally altered rocks or meteoric water. Lavas andtuffs erupted over nearly the next 1.5 million years recovered gradually in 18O indicating thatinfusions of magma from greater depths successively augmented the upper-crustal magmachamber; by the time of the climactic Lava Creek Tuff eruption at 639±2 ka, 18O had recovered tovalues in quartz as high as 6.8‰. Formation of the Yellowstone caldera, however, was againaccompanied by massive assimilation of low- 18O materials; quartz in early intracaldera lavasdrastically reduced in 18O to values as low as ~1‰. Since then, 18O in intracaldera lavas and tuffshas again gradually recovered with successive additions of high- 18O magma; during theintracaldera eruptions of ~170,000-70,000 years, 18O in quartz has increased irregularly from ~4.6to 5.1‰.

Trace-element compositions of the Central Plateau Member lavas became progressivelymore evolved with time (Hildreth and others, 1991; Vazquez and Reid, 2002). Vazquez and Reid(2002) showed that zircons in the Central Plateau Member lavas have 238U/230Th model ages thatgenerally range from their eruption ages back to about 200 ka—compared to U-Pb zircon data fromthe Lava Creek Tuff that show its magma to have been generated <100,000 years before thecaldera-forming eruption (Bindeman and others, 2001). Thus the Central Plateau Member lavascan be regarded as representing an incrementally augmented body of upper-crustal rhyolitic magmabeneath the Yellowstone caldera. Furthermore, these late postcaldera lavas show evidence ofprogressive differentiation throughout the time of their eruption. The ratio Rb/Sr, for example,generally increases with decreasing ages of the flows. Vazquez and Reid (2002) modeled mineraland whole-rock trace-element chemistry as being consistent with the sequence of lavas postdatingformation of the Yellowstone caldera representing at least 40% sanidine-dominated fractionalcrystallization of the residual magma body. This magma, initially low in 18O, was augmentedprogressively by infusions from deeper magma. Through most of this time, in the Vazquez andReid model, the lavas erupted intermittently while sanidine and other major phenocryst phasesfractionated progressively to accumulate below in a crystal-liquid mush but zircon did not. In fact,it may well be that any preexisting zircon and plagioclase were resorbed in the residual magma.After about 120 ka, new zircon crystals grew in magmatic liquid separated from the underlyingmush for as long as 50,000 years before eruption to produce the youngest of the Central Plateaulavas, the Pitchstone Plateau flow.

The successive eruption of these voluminous, progressively more evolved lavas from asingle magma body suggests a model in which magma is augmented incrementally but slowlycrystallizes and differentiates in a postcaldera Yellowstone magma chamber while the chamber roofis progressively cooled from the top by hydrothermal circulation. A mixture of magmatic liquidand crystals is envisioned to remain in the chamber until fractures driven by regional tectonicextension in the cooling chamber roof become large enough to extend downward toward the levelof magma storage. These fractures then enable the intrusion of rhyolitic dikes, allowing the magma

23

to extrude and erupt to the surface. The intrusions heat the chamber roof, reducing its elasticstrength, sealing and inhibiting further development of large, deep fractures, and tending to retardfurther eruption. Only as hydrothermal circulation again cools the rocks of the chamber roof canfractures of sufficient size again start to develop and promote renewed tapping of liquid from themagma chamber, renewed diking, and surface eruption. With time, magmatic liquid in the chambercontinues to differentiate, but smaller volumes remain sufficiently liquid to allow mobilization andeventual eruption. Thus, the rhyolites erupt in episodes of slightly more chemically evolvedcompositions. This model implies that any magmatic increments introduced into the upper-levelchamber between ~170 and 70 ka intruded at a rate lower than the rate of differentiation within themagma body.

Renewal of large rhyolitic eruptions within the Yellowstone caldera would present twomajor classes of hazards, explosive pyroclastic ejections and effusive lava eruptions. The initialphases of an eruption would undoubtedly involve pyroclastic eruption through an opening vent.Pyroclastic deposits probably lie buried beneath each of the large Central Plateau Member flowsbut remain mostly unexposed. Their distal equivalents, however, have been recognized insedimentary basins surrounding Yellowstone, perhaps including ash beds designated HebgenNarrows and Natural Trap by Izett (1981). Pumiceous ash mapped as the tuff of Cold MountainCreek may also represent pyroclastic materials erupted in vent-opening phases of one or more ofthe Central Plateau Member lavas. Burial by pumice lapilli and ash from such vent-openingeruptions could be widespread and locally as thick as many meters. Pyroclastic flows coulddevastate areas of many tens of square kilometers. Subsequent lava extrusion could last manyyears, covering areas as great as 350-400 km2 to thicknesses of tens or hundreds of meters andvolumes of 5 to more than 50 km3. Because such voluminous rhyolitic lava flows areunprecedented in the global historical record and their extrusion rates may be considerably greaterthan those of observed smaller flows, the actual length of time that an individual eruption mightcontinue is uncertain. Certainly no infrastructure in areas affected by such lavas would survive.

Note should also be taken here of the tuff of Bluff Point, the only voluminous ash-flow tuffof the Central Plateau Member; its eruption resulted in formation of the 10-km caldera nowrepresented by the West Thumb. The Bluff Point eruption was characterized by ejection, not onlyof pumice, ash, and small lithic fragments, but also of vitrophyric lava blocks as large as 1 m andpossibly involved copious amounts of water from a predecessor of Yellowstone Lake. Based onthis example, it is reasonable to suggest that any future voluminous rhyolitic eruption within orimmediately adjacent to Yellowstone Lake might have the potential to produce an explosivepyroclastic eruption, much smaller in volume than a major caldera-forming event like the LavaCreek Tuff but nevertheless capable of producing high-speed flows of hot ash, rhyolitic debris, andvolcanic gas that could affect virtually all of the Yellowstone Plateau area and produce regionallywidespread finer downwind ash.

Secondary hazards associated with large intracaldera rhyolitic eruptions could includewildland fires, disturbed drainages, and possibilities of catastrophic flooding, and they could beespecially severe and widespread given their great volumes. In particular, if a large-volume lavaflow should displace a considerable volume of Yellowstone Lake, overflow could inundate largeparts of the Yellowstone River drainage. Other large drainages could be impounded by lava, thenreleased as catastrophic floods overtopped or undermined the lava dams.

Estimating probabilities for one or more large intracaldera rhyolitic eruptions involvesconsiderable uncertainty. Considering that 17 known voluminous lava eruptions and 2 moderatelylarge pyroclastic eruptions are documented and that additional flows probably are buried within thecaldera, the raw probability might be estimated on the basis of approximately 25 events betweenabout 170,000 and 72,000 years ago, an average intereruptive period of about 3900 years duringthat time. Alternatively, if it is assumed that the probability of such an eruption has remained

24

uniform from ~170 ka to the present, the average period between eruptions would be 6800 years,the latest of which was ~72,000 years ago, or an annual probability of 1.5x10-4. Clearly, however,such calculations are misleading. The voluminous eruptions were highly episodic, with numerousevents occurring within periods of a few thousand years and longer periods of ~12,000 to 38,000years ensuing without eruption and about 72,000 years since the latest of them. Furthermore,generally decreased numbers of eruptions in successive episodes, a doubling of the post-eruptiveinterval compared to the longest previous intereruptive intervals, and the fact that the latest of thefive eruptive episodes was possibly a single voluminous event might even suggest that therepeating cycle has ended.

The foregoing calculated probabilities are based implicitly upon the assumption that theprobability of an eruption occurring within a given time does not depend upon the time elapsedsince the last previous eruption (statistically a Poisson, or simple exponential process). Because thelarge intracaldera eruptions of about the past 170,000 years are clustered in time, however, a moremeaningful probabilistic calculation might be based upon a different statistical basis, a mixedexponential process (Cox and Lewis, 1966; Nathenson, 2001). Because of the isotopic evidence forreorganization of the magma at about 200 ka and distinct changes in eruptive behavior andpetrologic characteristics of intracaldera lavas since about 170 ka, only the time since 170 ka isconsidered in the following analysis. (A complete discussion of this analysis and the relevantfigures is given in appendix 3; only the important conclusions are summarized here).

The chronology for all known Yellowstone intracaldera eruptions since 170 ka is shownversus event number in figure 27. Because of uncertainties in the ages of these eruptions, thisfigure shows three possible models for the ages for individual lava flows within each episode.Intereruptive intervals during certain periods are short, averaging less than 8 ky, and may be muchless than that. There are several longer interruptions of activity: one is 38±4 ky (thousand years),another is 30±6 ky, and two others are 12±6 ky. The resulting probability distribution (fig. 28)illustrates this behavior; the mixed exponential model curve matches the data well.

Using the mixed exponential probability distribution, the conditional probability that anintracaldera lava eruption will occur in the next year (fig. 29) decreases as the time since the lasteruption increases, changing very little after about 20 ky. The mixed exponential probability of anintracaldera rhyolitic lava eruption in the next year, ~72 ka after the last eruption, is equivalent toan annual rate of 5x10-5 or a recurrence rate of 1 in 20,000 years.

In addition to the probability of an eruption occurring in a given time, an importantconsideration is its likely size. Figure 30 shows the probability of a lava eruption within the calderahaving a volume greater that any particular value, based upon volumes estimated for the 19 knownlate intracaldera lava flows or domes (table 2). The data are logarithmic with volume over most ofthe range, with large-volume lava flows more probable in the caldera than smaller ones. Theprobability that the next intracaldera lava flow will have a volume greater than 5 km3 is about 60%;the probability of a volume greater than 10 km3 is about 40%.

If a large intracaldera rhyolitic eruption were to occur, there would almost certainly besignificant premonitory indications. Seismic swarms would probably be concentrated in any areaof diking as magma rose from the magma chamber toward the surface. Such seismicity wouldmost likely embrace the entire thickness of the brittle part of the crust above the chamber and mightwell display a pattern of decreasing focal depths with time. Furthermore, an initial intrusive phasewould probably be indicated by a linear array of epicenters although they might or might notcentralize toward a smaller area before the opening of an eruptive vent. Surface deformationshould be readily detectible, and the volume of localized inflation—distinguished from such areallygeneralized inflation or deflation as measured within the caldera in recent decades—would likelypresage the arrival of lava at the surface. Vent opening might well involve the local ground watersystem to produce phreatic or phreatomagmatic explosions. Initial pyroclastic eruption could

25

extend to stratospheric heights, and its duration and magnitude might be considered as indicative ofthe volume of possible subsequent lava extrusion.

Small rhyolitic eruptionsThe possibility of future rhyolitic lava eruptions in the Yellowstone system involves two

more or less distinct scenarios. In addition to the large intracaldera eruptions just discussed,relatively small eruptions (a few cubic kilometers or less) might occur either within or outside theYellowstone caldera.

Most of the small postcaldera eruptions of rhyolitic lava have occurred at locations outsidethe Yellowstone caldera. Of the 13 such eruptions known (fig. 10), all but one vented north of thecaldera, either in the Norris-Mammoth corridor or between the Madison River and the southernGallatin Range. A single postcaldera rhyolitic lava flow erupted south of the caldera, near thesouth entrance to Yellowstone National Park (fig. 10). Most of these extracaldera rhyolites arevolcanic domes with diameters of 0.5-2.5 km and volumes less than 0.5 km3. Only four are largerflows that cover between 2 and 20 km2 or have volumes as great as 4 km3.

Several of the small rhyolites are intimately associated with basalt that occurs both asquenched magmatic inclusions and larger masses within the rhyolite; the largest of these are theGardner River and Grizzly Lake rhyolite-basalt mixed-lava complexes. Two rhyolites, theObsidian Cliff and Crystal Spring flows, completely lack phenocrysts, indicating that they wereemplaced at or above their liquidus temperatures; the Crystal Spring flow also contains abundantquenched magmatic basaltic inclusions near its base. These observations suggest a role for basalticmagma in mobilizing the rhyolitic magma for intrusion to shallow depths. Geologic, geochemical,and isotopic data show that the extracaldera rhyolites erupted from individual magma sources thatwere related to local basaltic magmas but unconnected to the large rhyolitic magma chamberbeneath the Yellowstone caldera (Hildreth and others, 1984; Hildreth and others, 1991).

Although individual small rhyolitic eruptions have vented from several locations around theYellowstone caldera, by far the greatest number erupted within the Norris-Mammoth corridor. Allof those younger than about 170,000 years (table 1), equivalent in age to the youngest intracalderarhyolites, are located either in the Norris-Mammoth corridor or near the South Entrance toYellowstone National Park, within 25 km of the caldera margin. Unlike the major normal faultzones such as the Teton, Sheridan, and Hebgen Lake zones (fig. 11), faults in the Norris-Mammothcorridor are generally of small to moderate displacement. By contrast, the principal fault alongwhich the east side of the Gallatin Range is relatively uplifted, had a net late-Cenozoic stratigraphicthrow of about 2 km. Nevertheless, it has been inactive during postglacial time or longer. It wassuggested by Pierce and others (1991), that magmatic diking along the Norris-Mammoth corridorpostdating the Lava Creek Tuff, both rhyolitic and basaltic, effectively takes up a significantproportion of the extensional strain that otherwise would be expressed in normal displacement onthe fault zone bounding the east side of the Gallatin Range.

Two distinct types of primary hazards might be associated with a small rhyolitic eruption inthe Yellowstone area. The initial vent-opening phase of such an eruption would almost certainly bean explosive ejection of rhyolitic pumice. Most of the coarser pumice would be deposited close tothe vent, but finer ash probably would be dispersed downwind over distances of many kilometers.Thicknesses of the downwind deposits could range from several meters to a few centimeters or less.In addition, an explosive phase could generate pyroclastic flows, high-speed mixtures of hot ashand gas capable of devastating areas within several kilometers of the erupting vent. Ash fallscould cover large areas and cause damage to structures, power lines, and machinery by loadingthem, especially if such an eruption should occur under wet atmospheric conditions. Watersupplies also could be damaged by substantial ash falls. If pyroclastic flows occurred, they wouldbe highly dangerous and destructive to the local environment or to any structures or people in the

26

vicinity of the eruption. Such initial pyroclastic eruptions would likely last between a few hoursand a few weeks but quite likely would be succeeded by extrusion of viscous rhyolitic lava to forma dome or small lava flow. Areas up to several square kilometers could be covered by lava tens ofmeters thick. Although the immediate area affected by the primary hazards of deep burial bypumice, ash, or lava might be relatively small, the slow extrusion of lava would be likely tocontinue for many months or even years. Structures would not be able to withstand the lavaadvance, and any nearby roads or other infrastructure would certainly be affected. Advance of theviscous rhyolitic lava would be much slower than for a basaltic eruption; most facilities could besafely evacuated, and some of them, as well as vehicles, could be relocated during advance of atypical flow front.

Secondary hazards from a small rhyolitic eruption would be likely to include fires started bycontact with the extruding lava. Other secondary hazards might include flooding if drainages weretemporarily blocked or subsequently released suddenly. Because of the relatively slow advance ofsuch lavas, however, effective mitigation of such effects might be somewhat less difficult than for arapidly advancing basaltic lava flow.

The probability of a small rhyolitic eruption outside the Yellowstone caldera is quite low.Fourteen have occurred in postcaldera time (Christiansen, 2001; Spell and others, 2004) with anaverage intereruptive period of about 46,000 years. During the past 170,000 years there have been6 small extracaldera rhyolitic eruptions, equivalent to an average period of 28,000 years. Eruptionsin the past 170,000 years have been larger, typically producing flows of several km3, in comparisonto the domes of less than 0.5 km3 extruded earlier in the postcaldera history. Insufficient age dataexist to evaluate any possible episodic behavior of these youngest eruptions. The average annualprobability of a small extracaldera rhyolitic eruption in the Yellowstone Plateau volcanic field issmall. Calculated on the basis of the entire postcaldera period, it probably is on the order of 2x10-5;on the basis of the record of the past 170,000 years it would be about 4x10-5. Although lessprobable, a similar small rhyolitic eruption might also be possible within the Yellowstone calderaas well as in the extracaldera zone. Such extrusions, for example, formed the Douglas Knob andTrischman Knob domes on the Madison Plateau though these might be late-stage vent domesrelated to the much larger Bechler River flow.

It might be difficult in the early stages to distinguish typical premonitory seismic,deformation, or gas-emission indicators of a small rhyolitic eruption from those for a basalticeruption. Because of more viscous magma, uplift and tilting of the ground surface above a zone ofintensifying swarm seismicity might likely be more pronounced for a rhyolitic eruption. Also, assuch seismic signals are typically associated with subsurface separation of a gas phase from themagma, the time required for any relatively deep long-period seismicity that might develop couldbe longer for a viscous rhyolitic magma than for basaltic magma. Detecting any such signalsmight, however, require instrumentation sensitive to a broader frequency range than is currentlyavailable at Yellowstone, such as borehole strain meters.

Large caldera-forming eruptionOf all the possible eruptive hazards that might occur in the region of Yellowstone National

Park, by far the least likely is that of another major caldera-forming pyroclastic eruption of 100 km3

or greater. Three such events have occurred in about the past 2 million years, each associated witha cycle of precaldera and postcaldera rhyolitic volcanism lasting on the order of a million years. Inthe Island Park area, west of the 639±2-ka Yellowstone caldera, the older rhyolitic source areashave subsequently produced basaltic lava eruptions. In contrast, contemporaneous basaltic magmassurround the Yellowstone caldera, but none have erupted within the caldera. This pattern stronglysuggests that the crust where rhyolitic magma chambers existed during the previous two majorcaldera-forming eruptions and their associated rhyolitic volcanism has cooled and solidified

27

sufficiently to fracture and allow basaltic magmas to intrude from below, precluding the possibilityof large volumes of eruptible rhyolitic magma remaining there. However, the great heat flowrepresented by the massive long-lived hydrothermal circulation system of Yellowstone (Fournier,1989) as well as significant delays in seismic-wave travel times and wave attenuation imaged in theshallow crust beneath the Yellowstone caldera (Benz and Smith, 1984; Miller and Smith, 1999;Husen and others, 2004) strongly suggest the continued presence of magma. What remain mostuncertain are (1) the percentage of melt in the remaining, partly crystallized magma, (2) its degreeof interconnection, and (3) its potential eruptibility. The more than 600 km3 of highly differentiatedmagma that has erupted as lava flows within the caldera between ~170 and 72 ka represents avolume equivalent to a large caldera-forming eruption. Those eruptions perhaps partly degassedand depleted the magma sufficiently slowly without triggering voluminous pyroclastic eruptionsthat they may have rendered another major caldera-forming eruption from the present subcalderachamber unlikely.

Despite the seeming improbability of another large caldera-forming eruption atYellowstone, the severe consequences of any such eruption, if one should occur, render it importantto consider its potential hazards. The distribution of basalt around the Yellowstone caldera, asnoted above, would seem to restrict the area of potential accumulation of sufficient rhyoliticmagma for any such eruption to the vicinity of the Yellowstone caldera. It might seem thatpropagation of the Yellowstone hotspot toward the northeast through late Cenozoic time (comparefig. 4) could make a future caldera-forming eruption more likely to occur northeast of the presentcaldera center. Indeed, crustal density is notably low, hydrothermal heat flow high, and seismicwave speeds slow in the area of Hot Springs Basin, ~13 km northeast of the Sour Creek dome (fig.7). These characteristics, however, probably are related to crustal changes associated with the deepvapor-dominated hydrothermal system of that area (Miller and Smith, 1999). The pattern of youngpostcaldera rhyolitic magmatism of the Central Plateau Member suggests the possibility(Christiansen, 2001) that, since ~170 ka, rhyolitic magma could have been rejuvenated orremobilized beneath the western, or Mallard Lake, caldera segment in a manner analogous to therelation of the second-cycle Henrys Fork caldera to the older first-cycle Big Bend Ridge calderasegment (fig. 5). Although no particular site can be shown to be either more or less likely to be thesource of any possible future voluminous pyroclastic eruption, only a site within or immediatelyadjacent to the present caldera seems reasonable for a large pyroclastic eruption at Yellowstone.

Spell and others (2004) and Smith and Bennett (2006) have suggested that the basaltic andrhyolitic volcanism of the Norris-Mammoth corridor postdating the Lava Creek Tuff represents anevolving magma chamber that could presage a new caldera-forming eruption from that area. Thatscenario, however, seems to be contradicted by several lines of evidence. Geochemical andisotopic data, especially 18O, show that all three caldera cycles of the Yellowstone Plateau volcanicfield erupted from a continuously evolving shallow-crustal magma-storage system and that themagmas erupted in the Norris-Mammoth corridor are unrelated to them (Hildreth and others, 1984;Hildreth and others, 1991). Furthermore, volumes of rhyolite in that corridor are small, andbasaltic magma was available for eruption continuously during eruption of those rhyolites, asindicated by the undated Osprey Basalt that fills deep canyons cut into the youngest dated basalt(174±46 ka) as well as abundant quenched mafic inclusions in the 80-ka Crystal Spring rhyoliteflow. Additionally, whereas geophysical indicators of inflation and deflation of the Yellowstonecaldera are commonly coherent over the caldera area, the Norris-Mammoth corridor shows nocomparable indications of systematic deformation or seismicity. These factors argue against anylarge-scale rhyolitic magma-storage sufficient to promote and sustain a caldera-forming eruption inthe Norris-Mammoth corridor.

Although the probability of another major caldera-forming eruption at Yellowstone isextremely low, the hazards from any such eruption cannot be overstated. Once any such eruption

28

began, it probably would proceed quickly to a climax, which would last for many days. Anycaldera-forming eruption probably would begin at one or a few individual sites with pliniancolumns of hot volcanic ash rising into the stratosphere, and likely would spread to encompass aring-fracture system encircling the newly collapsing caldera. Lacking historical precedents forsuch an eruption, the length of time that an initial plinian phase might last is uncertain and could bea few hours to many weeks. Once the eruption developed past the initial phase, areas within andadjacent to the eruptive source would be overrun quickly by lateral flows of mixed volcanic ash,rocks, and gases at temperatures of several hundred degrees Celsius and speeds of a hundred km/hand greater as ash fall continued both within the source area and to great distances, even around theglobe. Ash-accumulation rates estimated for the well-studied Bishop Tuff of eastern California,roughly comparable to the Lava Creek Tuff though approximately half its volume, indicate that theBishop Tuff erupted continuously for several days (Wilson and Hildreth, 1997). A major caldera-forming eruption at Yellowstone probably would have a similar duration. All structures and livingthings within the areas overrun by ash flows would be destroyed and possibly vaporized. Theatmosphere above the eruption would be choked with pumiceous volcanic ash and lapilli,producing an extreme hazard to aviation, and the ash and associated fallout would quickly spreadfar beyond the eruptive area. Because of the complex patterns of upper-atmospheric windcirculation, there would likely not be a single simple plume of ash fallout in one downwinddirection, but large parts of western and central North America might well be buried by ash falls,ranging from many meters in thickness within tens of kilometers of the eruption to manycentimeters at continental distances (compare fig. 3). Fine ash and other particulates would circlethe globe.

The possible secondary hazards from a potential large caldera-forming eruption atYellowstone could themselves be greater than the primary hazards from other types of eruption.Suspended ash would continue to circulate in the upper atmosphere for many weeks, and could,together with volcanic gases associated with the eruption, affect global climate for several years.Reduced solar irradiation resulting from suspended ash and sulfuric-acid droplets could cool thelower atmosphere for several years, at least throughout the northern hemisphere and probablyglobally as well (Rampino and others, 1988). Other global climatic effects might last even longer,depending on the sensitivity of certain environments (Rampino, 1991), with possibly severeecological consequences (Rampino and Ambrose, 2000). In addition, the primary ash blanket overlarge regions, especially of the Western United States, would be eroded by rainfall and surfacedrainage to be redistributed from higher areas into lower-lying places, mantling some of them withmany meters of ash even where initial fall was only millimeters to centimeters thick. Manyreservoirs would be silted up, some of them becoming unusable; some dams might fail as a result ofovertopping by their reservoirs. Most of these secondary effects would be distributed over largeregions, far from Yellowstone; the devastating primary effects would far outweigh them in theproximal area.

Although the probability of a large caldera-forming eruption at Yellowstone is exceedinglysmall, it is exceedingly difficullt to make a defensible quantitative estimate of that probability. Asthere have been three such eruptions in about the past 2,100,000 years, there are only two inter-eruptive periods from which to gauge any additional possible interval between the third and apotential fourth such event. The first interval, between the Huckleberry Ridge (2.059±0.004 Ma)and Mesa Falls (1.285±0.004 Ma) caldera-forming events, was 774,000±5700 years. The secondinterval, between the Mesa Falls and Lava Creek (0.639±0.002 Ma) events, was 646,000±4400years. A statement, widely repeated in popular media, regards such eruptions as occurring atYellowstone “every 600,000 years” with the latest eruption having been “600,000 years ago”. Thisis commonly taken to imply that another such eruption is “overdue”. Such a statement isstatistically indefensible on the basis of the extrapolation of two intervals. (Even the simple

29

arithmetic average of the two intervals is 710,000 years, not 600,000 years). From the line ofreasoning outlined here, the probability of a fourth large caldera-forming event at Yellowstone canbe considered to be less than 1 in a million, below the threshold of hazards interest unless futurepremonitory phenomena, probably more severe than those recorded historically in caldera systemsaround the world (Newhall and Dzurisin, 1988), were to be recognized.

Premonitory indications of an impending major caldera-forming eruption at Yellowstonewould include intense swarm seismicity, perhaps localized near the site of an impending outbreak,but initial indications might not be greatly different from those for a smaller eruption. A magmabody large enough to sustain a major caldera-forming eruption would, however, be expectedeventually to respond as a whole. By the time seismicity and ground fracturing spread toencompass a larger area, equivalent to a potential caldera and perhaps to outline a ring-fracturesystem, major eruption might already be well underway. Based upon the geologic record of theLava Creek eruption, magma rising to shallow levels almost certainly would produce significantuplift of both a locus of possible outbreak and also a larger area of shallow rising magma. Quitepossibly ground fracturing would accompany intrusion to shallow crustal levels and might evenbegin to outline a ring-fracture system (compare to figure 6). Magmatic gases venting to theatmosphere before any ash or lava were to erupt, including CO2, various sulfur species (butespecially SO2), and halogens, might well be more evident and more copious from such a largeshallow magma body as it ascends into the brittle upper-crustal zone than would be expected for asmaller body that might lead to a single central-vent rhyolitic eruption. Although many of thespecific premonitory events for such an eruption might resemble precursors to a smaller rhyoliticeruption, the magnitudes would be expected to be correspondingly greater, there might begeophysical activity over an areally large source, and the course of events would be expected to bemore complex and of longer duration than for a smaller eruption.

Hydrothermal-explosion hazardsThe potential hazards and probabilities of future hydrothermal-explosion events in

Yellowstone National Park are considered in this section, based upon information summarizedearlier about previous such events at Yellowstone and elsewhere. (Also see appendices 1 and 2).

How often do they occur?For several reasons, it is difficult to estimate the frequency with which hydrothermal

explosions occur at Yellowstone National Park. First, there is no clear demarcation between aforceful geyser eruption that ejects mud and rock fragments and a more energetic explosion thatcreates a new crater. Second, no comprehensive catalog of explosion events has been compiledheretofore. Table 4 lists the results of a literature review that uncovered 26 examples ofhydrothermal explosions that have been described in the past 126 years. They refer to events inwhich a new feature was created or the size of an existing thermal feature was increasedsignificantly by explosive excavation. There are many other examples of geysering where mud androck were thrown from the vent, including an eruption near Wall Pool in the Upper Geyser Basinduring the Summer of 2006, while this report was being prepared. However, the catalog in table 4refers to events where the vent walls or conduit are shattered and fragments are dispersed outward.Nearly all of the explosions in table 4 occurred at thermal areas close to roads, where visitors aremost frequent and the likelihood of observation of an actual explosion or of recent deposits ishighest. It is likely that similar explosions are much more frequent than tabulated in table 4. Aconservative estimate would be that at least one rock-hurling explosion occurs every two years atYellowstone. Because these events are typically small and can occur any time of year, includingwhen few visitors are present, the likelihood of harm to any individual park visitor is small. For

30

example, while the annual probability of a small hydrothermal explosion within the park, assumingone eruption every two years, is 0.5, if visitors are generally absent from geyser basins at night andbetween December and March, then the annual probability of an explosion that could causepersonal injury falls to 0.5 x 0.67 (8 months/year) x 0.58 (14 hours/24-hour day), or 0.19.Moreover, explosions can occur away from trails and roads where people observe thermal features.When such factors are considered, the annual probability of an injury due to a small hydrothermalexplosion falls to somewhere between 0.01 and 0.1.

Larger events clearly represent a greater threat to park visitors and infrastructure. Mufflerand others (1971) and Morgan and others (in review) identified geologic deposits formed throughhydrothermal explosions within the past 14,000 years (table 3). Within the park, there are at least18 such features, all with diameters greater than 300 m. Two craters, Mary Bay and Turbid Lake,have diameters greater than 1500 m.

One can generalize the probability of hydrothermal explosions by plotting explosionmagnitude (crater size) versus frequency, simplified as annualized probability as shown in table 5.

In figure 31, these data plot as a straight line on a logarithmic plot of annualized probabilityversus the area of the newly formed crater (a proxy for eruption magnitude). Interpolation alongthis line implies that one could expect an explosion large enough to produce a crater 100 m indiameter every 200 years.

Such an analysis is consistent with the energetics of hydrothermal explosions. Browne andLawless (2001) estimated that the energy releases of even large explosions, producing craters 1000m across, represent only a few weeks to months of the heat flow from the system. Hedenquist andHenley (1985) produced a similar assessment for the prehistoric Waiotapu eruptions in NewZealand (craters ranging from 60 to 250 m in diameter). They concluded that the eruption thatformed the present Champagne pool (now 70 m in diameter) released the energy equivalent of 30hours of natural heat flux from that feature. Using such reasoning, Browne and Lawless (2001)advised that hydrothermal explosions should be relatively common events in long-lived geothermalareas and do not require unusual conditions.

Potential effectsAs stated above, small hydrothermal eruptions creating craters a few meters across are

unlikely to cause frequent injury to park visitors or damage to infrastructure. Larger explosions,however, are more energetic, and their effects could be correspondingly greater. We haveestimated that an explosion creating a 100-m-diameter crater might occur in Yellowstone aboutonce every 200 years. A series of similar-sized craters formed between one and two thousand yearsago in the Waiotapu geothermal system of New Zealand. As described by Hedenquist and Henley(1985), deposits from these explosions extend 5 to 20 times greater than the diameters of theirsource craters; thus, rock fragments might fly as far as 2000 m from a 100-m crater. Areas near theexplosion source would be buried beneath thick fragmental deposits. At the Okaro crater in NewZealand, deposits are 2 or more meters thick within 400 m of the originally 250-m-wide crater, andejecta deposits extend up to 1000 m from the vent (Hedenquist and Henley, 1985). Morgan andothers (in review) note that rock fragments were ejected at least as far as 3.5 to 4 km from the MaryBay crater rim during its period of explosive activity, ~13,600 years ago. Fragments from theIndian Pond explosion crater are found as far as 3 km from the vent (Morgan and others, 2003b).

Another potential hazard from hydrothermal explosions is the release of substantial amountsof toxic gases such as H2S and CO2. The topic of toxic gas is discussed later in this hazardassessment, but it is worth pointing out here that gas released during hydrothermal explosions atDieng, Indonesia, in 1979, killed 142 people. The gas was discharged from several adjacent areasthat had first ejected rocks and mud, forming several craters, the largest of which was 90 m wideand 100 m deep. At Yellowstone, there is no evidence that any hydrothermal explosions have been

31

accompanied by significant toxic-gas release. On the other hand, the Yellowstone caldera is aprolific source of CO2 (Werner and Brantley, 2003), and it remains possible that explosions fromvapor-dominated regions, areas without a record of historic explosions, may release significantquantities of gas.

Precursory signalsPrior to 1989 (when it exploded), Porkchop Geyser exhibited a variety of characteristics that

could be taken to indicate a potential explosion. Its behavior changed from that of an irregulargeyser to a perpetual spouter (i.e., in constant geyser-like eruption), and the temperature of its deepfeeding waters increased (Fournier and others, 1991). With the spring emitting a roar audible from2 km away, few visitors in the vicinity would have failed to notice its presence. Today, as then, ageyser behaving in such a manner would be monitored closely by YVO staff and collaborators.Other hydrothermal features in Yellowstone have experienced explosions after first displayingsimilar signs of instability. After the M=7.5 1959 Hebgen Lake earthquake, park geologist GeorgeMarler inventoried hundreds of features with altered behavior. One such feature, across theFirehole River from Biscuit Basin, originated as a set of newly formed fractures emitting steamwith apparent superheat, at ~95°C. A fumarole exploded from this site sometime in the earlyspring of 1963, forming a crater 2.7 x 4.9 m in size (Marler and White, 1975) that was subsequentlynamed Seismic Geyser. Currently, the feature is a non-erupting hot spring. Hydrothermalexplosions following earthquakes would be fundamentally unpredictable.

The Dieng, Indonesia, hydrothermal explosion of February 20, 1979 was preceded forseveral days by shallow seismic swarms (Allard and others, 1989). During the three hours prior tothe beginning of the explosions, three earthquakes were felt by nearby villagers 3 km distant fromthe explosion source. Once initiated, the explosive activity lasted for about 2 days. Other thermalfeatures at Dieng, as elsewhere, have exploded without conspicuous precursory activity (Allard andothers, 1989). White (1955) described an eruption at the Lake City Hot Springs northeasternCalifornia in March 1951 that dispersed 300,000 tons of mud, greatly modifying 20 acres of hot-spring area and dispersing fine dust at distances more than 6 km from the center. The activityoccurred without any warning to local residents and farmers. Hydrothermal explosions occurnearly every year from hot springs and geothermal wellheads in the Taupo Volcanic Zone of NewZealand. In particular, Rotorua, a developed geothermal area and tourist attraction, has producedscores of such eruptions over the past 100 years without obvious precursors (Scott and Cody, 2000;Browne and Lawless, 2001).

Clearly, some hydrothermal eruptions can initiate quickly, including but not limited to thosetriggered by earthquakes. Others seem to follow years of anomalous activity of hot springs andfumaroles. Because hydrothermal explosions typically appear to be shallow-rooted phenomena,initiating close to the surface, geophysical signals are most likely to affect localized areas withinonly a few hundred meters of the actual explosion site. At most volcanoes and hydrothermalsystems, including Yellowstone, the station spacing of seismometers, tilt meters and othermonitoring equipment will only rarely yield sufficient information to detect such precursoryactivity. Local monitoring networks with multiple sensors at a geyser basin might be sufficient todetect precursors such as ground tilting or subtle tremor, but as no such experiment has yet beenundertaken, this supposition remains unproven.

Where are hydrothermal explosions most likely to occur?Both historic (table 4 and fig. 25) and prehistoric (table 3 and fig. 24) hydrothermal-

explosion craters are located predominantly in the Firehole River geyser basins, in and aroundYellowstone Lake, and in the southern part of the Norris-Mammoth Corridor. The preponderance

32

of historic explosion sites within the Firehole River and Norris basins may partly reflect greatervisitation and scrutiny of thermal features in those areas but also is consistent with liquid-dominated systems having near-surface temperatures near boiling being more likely to experienceexplosions than vapor-dominated areas that are not continuously saturated with liquid.

The north half of Yellowstone Lake and the adjacent area has long-been known as a site oflarge explosion craters and hosts an active liquid-dominated geothermal system (Muffler andothers, 1971; Morgan and others, 1977; Otis and others, 1977; Wold and others, 1977; Morgan andothers, 2003b). Besides the explosion craters listed in table 3, numerous small sublacustrineexplosion craters have been identified within linear fissure zones by high-resolution imaging(Johnson and others, 2003; Morgan and others, 2003b; Morgan and others, in press-a). Johnsonand others (2003) and Morgan and others (2003b) considered domal areas on the lake floor to bethe most likely sites of future hydrothermal explosions. These areas are manifested by gentlywarped areas of the lake floor having at most 30 meters of apparent uplift (Morgan and others, inpress-b). Hydrothermal alteration is inferred from the nature of seismic reflections; doming isinterpreted to have occurred as a result of hydrothermal pressurization beneath a sealed zone of lowpermeability resulting from precipitation of siliceous sinter or silicification (Morgan and Shanks,2005). Some of these domal areas are breached by vents, the formation of which would haverelieved any notable overpressure. Domes without any evidence of breaching were considered byJohnson and others (2003) as the most likely to experience future explosive activity.

Figures 24 and 25 identify a significant number of historic and prehistoric explosion craterswithin the acid-sulfate terrains that overlie vapor-dominated systems. As such, one cannot negatethe possibility that these areas too are susceptible to explosive activity and may present asignificant hydrothermal-explosion hazard. It remains possible, however, that many of theobserved hydrothermal-eruption craters in vapor-dominated areas, such as Roaring Mountain, FernLake, Hot Springs Basin, and Sulphur Hills, were formed at a time preceding present vapor-dominated activity. White and others (1971) noted that the Mud Volcano region showed evidencefor having earlier been a liquid-dominated system. Many inactive sinter terraces were apparentlyencroached upon by acid-sulfate mineralization, indicating a local increase in the depth to theboiling ground water table and cessation of surface flow of silica-saturated waters. Fournier (1989)noted similar evidence of a previous liquid-dominated system that precipitated siliceous sinter atHot Springs Basin, an area today characterized almost solely by acid-sulfate mineralization andwithout outflow of appreciable neutral-Cl waters (Allen and Day, 1935). Plausibly, the largeexplosion craters in these areas occurred prior to or during a transition from liquid- to vapor-dominated conditions.

Though Yellowstone’s silica-depositing geyser basins and acid-sulfate thermal areas are allknown to host hydrothermal-explosion features, the same cannot be said for calcium carbonatetravertine deposits in areas such as Mammoth Hot Springs, Terrace Spring, and Calcite Springs.Typically, the waters issuing in these areas are 20-30°C below the local boiling temperature and,therefore, pose less threat of destabilization. No evidence for previous explosions has been foundin these areas. The greatest hazards related to travertine-forming areas are buildup of toxic orunhealthy levels of CO2, as discussed later, and local subsidence of the ground surface due toformation of sinkholes in travertine areas. Unlike silica, the solubility of calcite increases withdecreasing temperature, so that migration of cold ground water, especially when it is acidic, cancreate caves and sinkholes. Such features are common at Mammoth Hot Springs and pose acontinual threat to park infrastructure.

In summary, two types of thermal areas at Yellowstone could potentially host futurehydrothermal explosions. The geologic and historic records show that the thermal areas with activedischarge of neutral, high-chloride waters, especially those hosting geysers, are most likely toexperience explosions. Vapor-dominated regions covered with acid-sulfate soils may also

33

experience explosions, though perhaps less regularly than areas with abundant shallow boilingground water. Travertine-forming regions do not appear to experience hydrothermal explosions.

Seasonal and long-term effects on hydrothermal explosionsFournier and others (1991) noted that the 1989 explosion of Porkchop Geyser corresponded

with the onset of a seasonal disturbance common in the Norris Geyser Basin (summarized by Whiteand others, 1988), typically occurring in late summer and characterized by increased heatdischarge, increased boiling, and chemical changes. The disturbance is akin to an undergroundgeyser eruption, where an increase in local heat and boiling migrates through the shallowsubsurface (White and others, 1988; Fournier and others, 1991). White and others (1988) speculatethat the disturbances are due to changes in the local ground water pressure, typically related toseasonal fluctuations. The temporal correlation of the Porkchop explosion and the basin-wideNorris disturbance suggests that hydrothermal explosions might be more likely during periods ofincreased heat discharge and seasonal or longer-term fluctuations in the pressure of the groundwater table. The feedback between ground water pressure and earthquakes was recently discussedby L. B. Christiansen and others (2005), who noted seasonal variations in seismicity beneathYellowstone Lake, increasing during periods of reduced lake levels. The observed drop in the levelof Yellowstone Lake in late summer and early fall lowers the hydrostatic head on sublacustrinehydrothermal vents, apparently sufficient to cause increased release of H2S-rich gas bubbles andentrainment of fine-grained sediments in upwelling hydrothermal fluids (Morgan and Shanks,2005).

The potential effects of longer-term variations in climate on stability of hydrothermalsystems can be inferred from the data presented by Muffler and others (1971) and Bargar andFournier (1988), showing fluid-inclusion filling temperatures in secondary inclusions that werehigher than present-day hydrostatic boiling temperatures. After departure of glacial ice from theYellowstone region, the hydrothermal system apparently remained at least temporarily superheatedrelative to the newly lowered hydrostatic pressure. Under such conditions, the Yellowstonehydrothermal systems might have been especially susceptible to explosion-triggering mechanismssuch as earthquakes or drainage of glacial lakes. The established chronology of largehydrothermal-explosion craters is insufficient to show that large explosions were more commonduring the very latest Pleistocene, shortly following ice melting. Recent work (Pierce and others,2002; Morgan and others, in review) suggests that most large hydrothermal explosions inYellowstone occurred at times that preclude a direct link to deglaciation.

Future climate variability, even in the absence of glaciation, might be expected to affect thestability of hydrothermal reservoirs. Drops in the water level of Yellowstone Lake or reduction inthe ground water table at the Norris or Firehole River Geyser Basins could increase the likelihoodof explosive hydrothermal activity in those regions. Many vapor-dominated reservoirs displayevidence for conversion from an earlier stage as a liquid-dominated system (White and others,1971; Moore and others, 2000). This evolution may occur due to explosive venting, reducedrecharge, or other potential mechanisms. As discussed above, the increase in boiling associatedwith such a conversion could result in explosions. It is possible that long-term drought andassociated reduced recharge could induce hydrothermal explosions in reservoirs in the process ofconverting from liquid-dominated to vapor-dominated conditions.

Hazard mitigationInfrastructure: Hydrothermal explosions at Yellowstone occur solely in geothermal areas.

Foremost among these areas are the geyser basins where silica-saturated, neutral, Cl-rich watersemerge at or close to their boiling point. The Upper and Lower Geyser Basins, Norris Geyser

34

Basin, West Thumb Geyser Basin, and the area within and adjacent to the north half ofYellowstone Lake are the areas most likely to experience hydrothermal explosions. Parkinfrastructure should be minimized in these areas, and new construction would best be concentratedaway from hydrothermal basins.

Documentation: Areas exhibiting superheat (water or fumarole temperatures above thelocal boiling point) should be monitored carefully by YVO and park staff. Subsequent to large, feltearthquakes, park staff should document changes to thermal features using a formal archival systemthat can be made widely available and accessible.

Monitoring: The hazard from small hydrothermal explosions is relatively low. Smallevents, usually producing craters less than two meters across, happen every few years and are onlyrarely witnessed. However, events forming craters over 50 to 100 m in diameter do occur on asocietally relevant timescale and could appreciably impact park infrastructure and visitor safety.Observational monitoring can be augmented by limited geophysical monitoring of grounddeformation and seismicity in selected geyser basins and around the northern half of YellowstoneLake and, perhaps, repeated geochemical sampling of hydrothermal fluids. Existing seismic andGPS stations are placed so as to minimize the “noise” from thermal features and maximize thesignal from deep earthquakes and magmatic processes. More local monitoring designed to “listen”to geyser basins would initially be useful as a tool to understand the dynamics of geothermal areason monthly to annual timescales. It is quite possible that precursors to some hydrothermalexplosions would become evident in such monitoring.

Gas-emission hazardsThere has been no systematic study of the concentrations of CO2 and H2S in the air at

Yellowstone. Nevertheless, because human death or serious illness from toxic gases in the air isexceedingly rare at Yellowstone, it is clear that gas concentrations are rarely at levels that couldcause severe illness. In general, the flat topography in most of Yellowstone’s thermal areas and thelight to moderate winds are sufficient to readily disperse hydrothermal gases. Reconnaissanceefforts to record the abundances of CO2 and H2S in thermal areas show that concentrations at waistlevel are rarely above 0.1% and 2 ppm respectively, even adjacent to fumaroles (Jacob Lowensternand Henry Heasler, unpublished data). Much higher and even lethal concentrations occasionallycan be found in cracks, holes, and locations within 10 cm of the ground surface; such air, however,generally is unlikely to be inhaled.

Though concentrations of toxic gases are generally low in Yellowstone, it is clear that theycan build up in valleys, caves, and tunnels and during windless conditions. The impact of gasconcentrations on humans living in the park was documented as early as 1883, when a woman wasovercome by vapors in the basement of one of the homes in Mammoth Hot Springs. The woman’sabsence was quickly noticed, and she was rushed into fresh air (Whittlesey, 1995, p. 66). Cavesand sinkholes near Mammoth Hot Springs are known to harbor relatively high concentrations ofCO2, which bubbles out of the local hot-spring waters. Poison Spring, Poison Cave, and theStygian Cave are all known as places where dead birds, insects, and small mammals have beenfound (Whittlesey, 1995, p. 68).

“Death Gulch,” is a well-known source of CO2 in the northeast section of YellowstoneNational Park, along a short tributary of Cache Creek, 3 km above its confluence with the LamarRiver. Temperatures in this area are well below boiling, but springs yield copious bubbles of gas,predominantly CO2, H2S, and methane (CH4) (Werner and Brantley, 2003). During his visit in1888, USGS geologist Walter Weed found the remains of six bears, an elk, squirrels, hares andnumerous butterflies and insects in the gulch (Weed, 1889). All were concluded to have diedthrough asphyxiation where CO2 and H2S had accumulated to toxic levels. One of the bears hadnot yet begun to decompose, and Weed concluded that the death had been very recent. Death

35

Gulch was visited in 1897 by T. A. Jaggar, who found the carcasses of seven grizzly bears and oneblack bear (Traphagen, 1904). Traphagen visited Death Gulch in 1903 and recorded more than10% CO2 in air near the bottom of the gulch and verified that gas emitted from openings in theground was mainly CO2 but contained over 10,000 ppm H2S. Flies placed into these openings diedwithin six seconds (Traphagen, 1904).

Yellowstone’s only documented human death by gas inhalation occurred in 1939 whenworkers were lowered in a bucket into a 26-foot-deep pit near the Yellowstone River at TowerJunction, dug to gather information about geological conditions relevant to nearby construction(Whittlesey, 1995, p. 67). In trying to rescue a fellow construction worker who was overcome bythe vapors, another worker, Bill Nelson, lost consciousness and later died. Later tests revealed H2Sconcentrations of 200 to 400 ppm in the pit and CO2 levels of 20%. The combination wassufficient to cause another rescuer to lose consciousness after only a few short breaths. After thisincident, Bureau of Mines engineers also tested the air at Devils Kitchen at Mammoth Hot Springs.Their finding of 7% CO2 spurred them to recommend closing of that and other nearby caves tovisitation (Whittlesey, 1995). The caves have remained closed since then.

A more recent episode of toxic-gas inhalation occurred in March 2004, when five bisondied along a section of the Gibbon River adjacent to the Norris Geyser Basin (Heasler andJaworowski, 2004). The bison were all lying in similar positions, with their legs stifflyperpendicular to their bodies, and showed no signs of poor health prior to their deaths. Parkbiologists and geologists concluded that toxic gas inhalation was the most likely cause of death(Heasler and Jaworowski, 2004). Though the area is adjacent to thermal features rich in CO2 andH2S, dangerous concentrations of these gases were not found in ambient air at the time of discoveryof the animals, or since then. Consideration of the previous month’s weather records causedHeasler and Jaworowski (2004) to conclude that toxic levels of gas accumulated along the riverbottom during a brief time of cold weather and associated atmospheric inversion about one weekbefore the animals were found. It remains unknown whether a spike in outgassing from thehydrothermal system may have contributed to the probable gas accumulation.

Relevant examples of toxic volcanic or hydrothermal gas hazardsAs noted above, Yellowstone differs from most volcanoes in having very low emissions of

acid gases such as SO2 and HCl. A notorious hazard at some volcanic systems is buildup of CO2 indeep volcanic lakes, which can result in catastrophic degassing of the deep waters. Such eventshappened twice in Cameroon from Lakes Monoun and Nyos in 1984 and 1986, respectively. Thelatter event occurred when a landslide into the lake caused dissolved CO2 to be released rapidlyfrom solution, causing a cloud of CO2 to flow downhill from the crater lake and killing 1700 peopleand even more livestock (Sigurdsson, 1988). Following these events, engineers constructedvertical pipes in the lakes to ensure continual degassing of the deep gas-charged waters. TheCameroon Lakes are meromictic, meaning they are stratified into different layers that do notnormally mix, a consequence of the geometry of the lakes (deep and of small diameter), thetemperate climate, and the fact that CO2-rich water is denser than the fresh water above.Yellowstone Lake and the other lakes of Yellowstone National Park are dimictic; their deep andshallow waters mix twice per year, in the spring and fall (Thompson and others, 1998).Additionally, hot springs on the floor of Yellowstone Lake tend to increase the temperatures ofbottom waters, enabling mixing as the warm waters rise. Because of the continual mixing, it ishighly unlikely that dissolved CO2 concentrations in the deep waters of Yellowstone lakes couldincrease to levels sufficient to cause catastrophic degassing.

Though CO2 is unlikely to be released suddenly due to lacustrine processes, it remainspossible that dangerous amounts of this gas could be released by an explosion from thehydrothermal system. A precedent for such an event occurred in the Dieng Plateau, a volcanic area

36

in Indonesia known for accumulations of CO2 in depressions and hollows, as well as for frequenteruptions of mud from volcanic craters. In 1979, 134 people were killed while attempting to escapefrom the area following several days of earthquakes, hydrothermal explosions, and anaccompanying mud flow that blocked access to the village (Allard and others, 1989). Anotherdeath by CO2 poisoning at Dieng occurred in March 1992 (Global Volcanism Network, 1992).

Several deaths have been attributed to CO2 poisoning at Mammoth Mountain, a ski resort ina volcanic region of eastern California. Carbon dioxide discharge was observed there in the early1990s and resulted in several tree-kills that covered 70 hectares. High concentrations of CO2 havebeen measured there in depressions in the snow around trees and adjacent to buildings (Sorey andothers, 1996). In 1998, a skier is believed to have died as a result of CO2 poisoning after fallinginto a snow well adjacent to a U.S. Forest Service restroom (Hill, 2000). In March 2006, threeadditional gas-related deaths occurred at Mammoth Mountain as members of the MammothMountain ski patrol made repairs to fencing that surrounds a near-boiling fumarole high on themountain (Becerra, 2006). Although the area is normally roped off during the winter because thefumarole melts a hole in the snow—a hazard for skiers—a heavy snowfall partially obscured thefence. During repairs, two individuals fell into the hole and quickly succumbed to the CO2 gas.Another team member died during a subsequent rescue attempt. The tragedy was clearlyengendered by the unusually thick snow pack, but the ultimate hazard was the fumarole itself andits CO2 emissions.

Hazard mitigationGiven the prodigious CO2 and H2S emissions at Yellowstone, the former estimated at

45,000 tons per day (Werner and Brantley, 2003), it is clear that toxic gases pose a potential hazardto park visitors. Normally, the gases dissipate quickly into the atmosphere, and only rarely arevisitors present where gas concentrations can cause harm. Nevertheless, incidents such as those atDieng and Mammoth Mountain are possible at Yellowstone. Also, unless safety measures areutilized, incidents similar to the one that caused a worker’s death at Tower Junction could occuragain. It is important that National Park staff working in suspect areas be made aware of thehazards from toxic gases, carry appropriate equipment to detect anomalous gas concentrations, andfollow recommended protocols. Caves near thermal areas should continue to be closed to visitors,including park staff. As is park policy, visitors in thermal areas should stay on maintainedpathways and away from thermal features and enclosed areas where dense gases can accumulate.

Because so few data on gas concentrations in thermal areas have been collected, werecommend further reconnaissance studies, including continuous monitoring to assess the effects ofwind, air temperature, and other climatic factors on gas concentrations in and around specificthermal areas. These types of studies would aid in determining whether occasional gas surges fromthermal features produce high concentrations of toxic gas in the surrounding air and what hazardsuch events might pose to park staff and visitors. Monitoring of buildings, particularly basements,might also be justified in areas of possible toxic-gas accumulation.

ConclusionsYellowstone National Park is the site of a large, active, and integrated tectonic, magmatic,

and hydrothermal system. Potential hazards related to possible future activity of this system, asevaluated in this report, vary greatly among the various possible types of future activity.

Earthquake hazards, although they are among the most commonly recurring hazards, are notevaluated explicitly in this preliminary open-file version of the assessment. Their assessment,however, is planned for a forthcoming, more complete version. Of the remaining hazards discussedhere, the most likely to occur are hydrothermal explosions, with an average annual probability from

37

as high as 0.5 (equivalent to an average recurrence of 2 years) for small explosions to perhaps5x103 (an average recurrence of 200 years) for explosions large enough to form a 100-m-diametercrater. Hydrothermal explosions that might result in potential risks to people have a probability nogreater than 0.1 per year (an average recurrence of 10 years) for small explosions.

Potential for a volcanic eruption is much lower than that for hydrothermal explosions.Probabilities range from about 6x10-5 per year (an average recurrence of 16,000 years) for a basalticeruption in the region to about 5x10-5 (average recurrence of 20,000 years) for a large rhyolitic lavaeruption within the Yellowstone caldera or 2x10-5 per year (average recurrence of 50,000 years) fora small extracaldera rhyolitic eruption. The probability of a large caldera-forming eruption is verymuch smaller than any of these but is not readily quantifiable; it probably is less than 10-6. Theaggregate annual probability of any volcanic eruption occurring from the Yellowstone magmaticsystem is ~1x10-4, an average recurrence of 10,000 years.

Continued monitoring by YVO is likely to enable recognition of premonitory indicationsbefore any volcanic eruption. There are likely to be few if any indications of an impendinghydrothermal explosion although certain conditions, such as major earthquakes or seasonal or long-term lowering of local water tables, tend to favor the occurrence of hydrothermal explosions. TheYVO monitoring system should at least provide rapid information about any such event once itoccurs. Toxic gas releases, especially of CO2 or H2S, would probably be recognized only after theyoccur, mainly by indirect evidence such as dead or distressed plants or animals.

Advance preparation for responses to potential hazards from future volcanic eruptions orhydrothermal explosions can help mitigate their effects on Yellowstone National Park and its staffand visitors. Despite the low probability of a future volcanic eruption at Yellowstone, thehazardous consequences possible from even a small eruption render important the preparation ofplans for appropriate responses to any such future event.

AcknowledgmentsWe appreciate the help of Jeff Cross and Ralph Taylor, both members of the Geyser

Observation and Study Association (GOSA), who provided documentation of some of the historichydrothermal explosions at Yellowstone. Nicole Nastanski and Terry Spell of the University ofNevada, Las Vegas and John Obradovich provided unpublished 40Ar/39Ar age data for Yellowstonevolcanic rocks. Andy Calvert provided unpublished 40Ar/39Ar dates of Yellowstone rhyolites andadvised us more broadly on issues related to geochronology. Tom Olliff and John Varley ofYellowstone National Park were supportive of this effort and helpful throughout its planning andpreparation. Kathryn Flynn assembled much of the geographic, monitoring and scientific data usedto construct the figures. Shaul Hurwitz, Bill Evans, Pat Shanks, and Deborah Bergfeld of theUSGS reviewed portions of this report relevant to hydrothermal and gas hazards. Wes Hildreth andDave Hill provided detailed technical reviews of the entire document.

References CitedAllard, P., Dajlevic, D., and Delarue, C., 1989, Origin of carbon dioxide emanation from the 1979

Dieng eruption, Indonesia: Implications for the origin of the 1986 Nyos catastrophe: Journalof Volcanology and Geothermal Research, v. 39, p. 195-206.

Allen, E. T., and Day, A. L., 1935, Hot springs of the Yellowstone National Park: Washington,D.C., Carnegie Institution of Washington Publication 466, 525 p.

Anders, M. H., Geissman, J. W., Piety, L. A., and Sullivan, J. T., 1989, Parabolic distribution ofcircumeastern Snake River Plain seismicity and latest Quaternary faulting: Migratorypattern and association with the Yellowstone hotspot: Journal of Geophysical Research, v.94, p. 1589-1621.

38

Bacon, C. R., 1983, Eruptive history of Mount Mazama and Crater Lake caldera, Cascade Range,U.S.A.: Journal of Volcanology and Geothermal Research, v. 18, p. 57-115.

Bargar, K. E., and Fournier, R. O., 1988, Effects of glacial ice on subsurface temperatures ofhydrothermal systems in Yellowstone National Park, Wyoming: Fluid-inclusion evidence:Geology, v. 16, p. 1077-1080.

Baxter, P. J., 2005, Human impacts of volcanoes, in Marti, J., and Ernst, G. G. J., eds., Volcanoesand the Environment: Cambridge, Cambridge University Press, p. 273-303.

Bebbington, M. S., and Lai, C. D., 1996, On nonhomogeneous models for volcanic eruptions:Mathematical Geology, v. 28, p. 585-600.

Becerra, H., 2006, Ski resort cited in toxic fume deaths, Los Angeles Times, October 7, 2006.Benz, H. M., and Smith, R. B., 1984, Simultaneous inversion for lateral velocity variations and

hypocenters in the Yellowstone region using earthquake and refraction data: Journal ofGeophysical Research, v. 89, p. 1208-1220.

Bindeman, I. N., Valley, J. W., Wooden, J. L., and Persing, H. M., 2001, Post-caldera volcanism:In situ measurement of U-Pb age and oxygen isotope ratio in Pleistocene zircons fromYellowstone caldera: Earth and Planetary Science Letters, v. 189, no. 197-206.

Browne, P. R. L., and Lawless, J. V., 2001, Characteristics of hydrothermal eruptions, withexamples from New Zealand and elsewhere: Earth-Science Reviews, v. 52, p. 299-331.

Bryan, T. S., 1995, The Geysers of Yellowstone, Third Edition: Boulder, Colo., University Press ofColorado., 472 p.

Camp, V. E., and Ross, M. E., 2004, Mantle dynamics and genesis of mafic magmatism in theintermontane Pacific Northwest: Journal of Geophysical Research, v. 109, no. B08204, p.doi:10.1029/2003JB002838.

Christiansen, L. B., Hurwitz, S., Saar, M. O., Ingebritsen, S. E., and Hsieh, P. A., 2005, Seasonalseismicity at western United States volcanic centers: Earth and Planetary Science Letters, v.240, p. 307-321.

Christiansen, R. L., 1979, Cooling units and composite sheets in relation to caldera structure, inChapin, C. E., and Elston, W. E., eds., Ash-flow Tuffs: Geological Society America SpecialPaper 180, p. 29-42.

———, 1982, Late Cenozoic volcanism of the Island Park area, eastern Idaho, in Bonnichsen, B.,and Breckenridge, R. M., eds., Cenozoic geology of Idaho: Idaho Bureau of Mines andGeology Bulletin 26, p. 345-368.

———, 1984, Yellowstone magmatic evolution: Its bearing on understanding large-volumeexplosive volcanism, in Explosive volcanism: Inception, Evolution, and Hazards:Washington, D.C., National Academy of Sciences, p. 84-95.

———, 2001, The Quaternary and Pliocene Yellowstone Plateau volcanic field of Wyoming,Idaho, and Montana: U.S. Geological Survey Professional Paper 729-G, 144 p.http://pubs.usgs.gov/pp/pp729g/.

Christiansen, R. L., and Blank, H. R., Jr., 1972, Volcanic stratigraphy of the Quaternary rhyoliteplateau in Yellowstone National Park: U.S. Geological Survey Professional Paper 729-B, 18p.

———, 1974, Geologic map of the Madison Junction quadrangle, Yellowstone National Park,Wyoming: U.S. Geological Survey Geologic Quadrangle Map GQ-1190, scale 1:62,500.

Christiansen, R. L., Foulger, G. R., and Evans, J. R., 2002, Upper-mantle origin of the Yellowstonehotspot: Geological Society of America Bulletin, v. 114, p. 1245-1256.

Cox, D. R., and Lewis, P. A. W., 1966, The statistical analysis of series of events: London,Methuen, 285 p.

39

Crone, A. J., 1987, Surface faulting associated with the 1983 Borah Peak earthquake atDoublespring Pass Road, east-central Idaho: Geological Society of America Abstracts withPrograms, v. 20, no. 7, p. 13.

Davis, C. C., 1930, New pool near Shoshone snowshoe cabin: Yellowstone Nature Notes, v. 4, no.4, July 1930.

Dewey, J. W., 1983, Instrumental seismicity of central Idaho: Seismological Society of AmericaBulletin, v. 77, p. 819-836.

Doser, D. I., 1985, Source parameters and faulting processes of the 1959 Hebgen Lake, Montana,earthquake sequence: Journal of Geophysical Research, v. 90, p. 4537-4555.

Doser, D. I., and Smith, R. B., 1983, Seismicity of the Teton-Southern Yellowstone Region,Wyoming: Seismological Society of America Bulletin, v. 73, p. 1369-1394.

———, 1989, An assessment of source parameters of earthquakes in the Cordillera of the westernUnited States: Seismological Society of America Bulletin, v. 79, p. 1383-1409.

Dzurisin, D., Savage, J. C., and Fournier, R. O., 1990, Recent crustal subsidence at Yellowstonecaldera, Wyoming: Bulletin of Volcanology, v. 52, p. 247-270.

Dzurisin, D., and Yamashita, K. M., 1987, Vertical surface displacements at Yellowstone caldera,Wyoming: Journal of Geophysical Research, v. 92, no. B13, p. 13,753-713,766.

Dzurisin, D., Yamashita, K. M., and Kleinman, J. W., 1994, Mechanisms of crustal uplift andsubsidence at the Yellowstone caldera, Wyoming: Bulletin of Volcanology, v. 56, p. 261-270.

Farrell, J., 2007, The time-spatial seismicity and development of a GIS database for theYellowstone region: Salt Lake City, University of Utah, M.S. thesis, 70 p.

Fournier, R. O., 1989, Geochemistry and dynamics of the Yellowstone National Park hydrothermalsystem: Annual Reviews of Earth and Planetary Sciences, v. 17, p. 13-53.

Fournier, R. O., Thompson, J. M., Cunningham, C. G., and Hutchinson, R. A., 1991, Conditionsleading to a recent small hydrothermal explosion at Yellowstone National Park: GeologicalSociety of America Bulletin, v. 103, p. 1114-1120.

Gansecki, C. A., Mahood, G. A., and McWilliams, M. O., 1996, 40Ar/39Ar geochronology ofrhyolites erupted following collapse of the Yellowstone caldera, Yellowstone Plateauvolcanic field: Implications for crustal contamination: Earth and Planetary Science Letters,v. 42, no. 91-107.

Global Volcanism Network, 1992, Sudden gas emission kills one person and hospitalizes twoothers: Bulletin of the Global Volcanism Network, April 1992, v. 17, no. 4,http://www.volcano.si.edu/world/volcano.cfm?vnum=0603-20=.

Handal, S., and Barrios, L. A., 2004, Hydrothermal eruptions in El Salvador: A review, in Rose, W.I., Bommer, J. J., Lopez, D. L., Carr, M. J., and Major, J. J., eds., Natural Hazards in ElSalvador: Geological Society of America Special Paper 375, p. 245-255.

Heasler, H., and Jaworowski, C., 2004, Geologic overview of a bison-carcass site at Norris GeyserBasin, March 2004: Yellowstone National Park Geology Report, 7 p.,http://www.nps.gov/yell/naturescience/upload/norrisbison.pdf.

Hedenquist, J. W., and Henley, R. W., 1985, Hydrothermal eruptions in the Waiotapu geothermalsystem, New Zealand: Their origin, associated breccias and relation to precious metalmineralization: Economic Geology, v. 80, p. 1640-1668.

Hildreth, W., Christiansen, R. L., and O'Neil, J. R., 1984, Oxygen isotopic study of theYellowstone Plateau volcanic field: Journal of Geophysical Research, v. 89, p. 8339-8369.

Hildreth, W., Halliday, A. N., and Christiansen, R. L., 1991, Isotopic and chemical evidenceconcerning the genesis and contamination of basaltic and rhyolitic magma beneath theYellowstone Plateau volcanic field: Journal of Petrology, v. 32, p. 63-138.

40

Hill, P. M., 2000, Possible asphyxiation from carbon dioxide of a cross-country skier in easternCalifornia: A deadly volcanic hazard: Wilderness and Environmental Medicine, v. 11, p.192-195.

Hobart, J. R., 1989, A Norris explosion crater update: Geyser Observation and Study AssociationTransactions, v. 1, p. 209-210.

Husen, S., and Smith, R. B., 2004, Probabilistic earthquake relocation for the Yellowstone NationalPark Region, Wyoming: Seismological Society of America Bulletin, v. 94, no. 3, p. 880-896.

Husen, S., Smith, R. B., and Waite, G. P., 2004, Evidence for gas and magmatic sources beneaththe Yellowstone volcanic field from seismic tomographic imaging: Journal of Volcanologyand Geothermal Research, v. 131, p. 397-410.

Hutchinson, R. A., 1985, Hydrothermal changes in the Upper Geyser Basin, Yellowstone NationalPark, after the 1983 Borah Peak, Idaho, earthquake, in Stein, R. S., and Bucknam, R. C.,eds., Proceedings of Workshop XXVIII on the Borah Peak, Idaho, earthquake: U.S.Geological Survey Open-file Report OF85-290, p. 612-624.

———, 1987, New thermal feature develops near Norris Geyser Basin: Yellowstone ParkMemorandum N3039(YELL), 2 p.

———, 1990, Thermal and seismic highlights of Yellowstone, 1986-87: Yellowstone NationalPark Archives, Box N-211, Folder 14.

Izett, G. A., 1981, Volcanic ash beds: Recorders of upper Cenozoic silicic pyroclastic volcanism inthe Western United States: Journal of Geophysical Research, v. 86, no. B11, p. 10200-10222.

Izett, G. A., and Wilcox, R. E., 1982, Map showing localities and inferred distributions of theHuckleberry Ridge, Mesa Falls, and Lava Creek ash beds (Pearlette family ash beds) ofPliocene and Pleistocene age in the Western United States and Southern Canada: U. S.Geological Survey Miscellaneous Investigations Series Map I-1325, 1:4,000,000.

Johnson, S. Y., Stephenson, W. J., Morgan, L. A., Shanks, W. C., III, and Pierce, K. L., 2003,Hydrothermal and tectonic activity in northern Yellowstone Lake: Geological Society ofAmerica Bulletin, v. 115, p. 954-871.

Keith, T. E. C., and Muffler, L. J. P., 1978, Minerals produced during cooling and hydrothermalalteration of ash-flow tuff from Yellowstone drill hole Y-5: Journal of Volcanology andGeothermal Research, v. 3, p. 373-402.

Klein, F. W., 1982, Patterns of historical eruptions at Hawaiian volcanoes: Journal of Volcanologyand Geothermal Research, v. 12, p. 1-35.

Lanphere, M. A., Champion, D. E., Christiansen, R. L., Izett, G. A., and Obradovich, J. D., 2002,Revised ages for tuffs of the Yellowstone Plateau volcanic field: Assignment of theHuckleberry Ridge Tuff to a new geomagnetic polarity event: Geological Society ofAmerica Bulletin, v. 114, p. 559-568.

Locke, W. W., and Meyer, G. A., 1994, A 12,000-year record of vertical deformation across theYellowstone caldera margin: The shorelines of Yellowstone Lake: Journal of GeophysicalResearch, v. 99, no. B10, p. 20,079-020,094.

Lowenstern, J. B., Christiansen, R. L., Smith, R. B., Morgan, L. A., and Heasler, H., 2005, Steamexplosions, earthquakes, and volcanic eruptions--What's in Yellowstone's future?: U.S.Geological Survey Fact Sheet 2005-3024, 6 p., http://pubs.usgs.gov/fs/2005/3024/.

MacLeod, N. S., Walker, G. W., and McKee, E. H., 1976, Geothermal significance of eastwardincrease in age of upper Cenozoic rhyolitic domes in southeastern Oregon, Second UnitedNations Symposium on Development and Use of Geothermal Resources, Proceedings:Washington, D.C., p. 465-474.

41

Mandavi, S., 2005, Toxicity, Hydrogen Sulfide, http://www.emedicine.com/EMERG/topic258.htm,updated 20 December 2005.

Marler, G. D., 1964, Effects of the Hebgen Lake earthquake of August 17, 1959, on the hot springsof the Firehole Geyser basins, Yellowstone National Park: U.S. Geological SurveyProfessional Paper 435-Q, p. 185-197.

———, 1973, Inventory of thermal features of the Firehole River Geyser Basins, YellowstoneNational Park: Washington, D.C., National Technical Information Service Report PB221289, 648 p.

Marler, G. D., and White, D. E., 1975, Seismic Geyser and its bearing on the origin and evolutionof geysers and hot springs of Yellowstone National Park: Geological Society of AmericaBulletin, v. 86, p. 749-759.

Meertens, C. M., and Smith, R. B., 1991, Crustal deformation of the Yellowstone caldera from firstGPS measurements: 1987-1989: Geophysical Research Letters, v. 18, p. 1763-1766.

Meyer, G. A., and Locke, W. W., 1986, Origin and deformation of Holocene shoreline terraces,Yellowstone Lake, Wyoming: Geology, v. 14, p. 699-702.

Milby, T. H., and Baselt, R. C., 1999, Hydrogen sulfide poisoning: clarification of somecontroversial issues: American Journal of Industrial Medicine, v. 35, no. 2, p. 192-195.

Miller, D. S., and Smith, R. B., 1999, P and S velocity structure of the Yellowstone volcanic fieldfrom local earthquake and controlled-source tomography: Journal of Geophysical Research,v. 104, no. B7, p. 15,105-115,121.

Moore, J. N., Adams, M. C., and Anderson, A. J., 2000, The fluid inclusion and mineralogic recordof the transition from liquid- to vapor-dominated conditions in The Geysers geothermalsystem, California: Economic Geology, v. 95, no. 8, p. 1719-1737.

Morgan, L. A., and Shanks, W. C., 2005, Influences of rhyolitic lava flows on hydrothermalprocesses in Yellowstone Lake and on the Yellowstone Plateau, in Inskeep, W. P., andMcDermott, T. R., eds., Geothermal biology and geochemistry in Yellowstone NationalPark: Bozeman, Mont., Montana State University Thermal Biology Institute, p. 31-52.

Morgan, L. A., Shanks, W. C., III, Lovalvo, D., Pierce, K. L., Lee, G., Webring, M., Stephenson,W. J., Johnson, S. Y., Finn, C., Schulze, B., and Harlan, S. S., 2003a, The floor ofYellowstone Lake is anything but quiet!: Yellowstone Science, v. 11, no. 2, p. 14-30.

Morgan, L. A., Shanks, W. C., Lee, G. A., and Webring, M., in press-b, Bathymetry, geology, andselected perspective views of the floor of Yellowstone Lake: U.S. Geological SurveyScientific Investigations Map.

Morgan, L. A., Shanks, W. C., Lovalvo, D., Lee, G., Pierce, K. L., Webring, M., Stephenson, W. J.,Johnson, S. Y., Finn, C., and Harlan, S. S., in press-a, The floor of Yellowstone Lake isanything but quiet: New discoveries from high-resolution sonar imaging, seismic reflectionprofiling, and submersible studies, in Morgan, L. A., ed., Integrated geoscience studies inthe Greater Yellowstone Geoecosystem: U.S. Geological Survey Professional Paper 1717.

Morgan, L. A., Shanks, W. C., and Pierce, K. L., in review, Hydrothermal processes above a largemagma chamber: Large hydrothermal systems and explosions in Yellowstone NationalPark.

Morgan, L. A., Shanks, W. C. I., Lovalvo, D. A., Johnson, S. Y., Stephenson, W. J., Pierce, K. L.,Harlan, S. S., Finn, C. A., Lee, G., Webring, M., Schulze, B., Duhn, J., Sweeney, R., andBalistrieri, L., 2003b, Exploration and discovery in Yellowstone Lake: Results from high-resolution sonar imaging, seismic reflection profiling and submersible studies: Journal ofVolcanology and Geothermal Research, v. 122, p. 221-242.

Morgan, P., Blackwell, D. D., Spafford, R. E., and Smith, R. B., 1977, Heat flow measurements inYellowstone Lake and the thermal structure of the Yellowstone caldera: Journal ofGeophysical Research, v. 82, p. 3719-3732.

42

Muffler, L. J. P., White, D. E., and Truesdell, A. H., 1971, Hydrothermal explosion craters inYellowstone National Park: Geological Society of America Bulletin, v. 82, p. 723-740.

Muir, J., 1888, Picturesque California and the Region West of the Rocky Mountains, from Alaskato Mexico: J. Dewing Publishing Co., 486 p.

Mullineaux, D. R., 1974, Pumice and other pyroclastic deposits in Mount Rainier National Park,Washington: U. S. Geological Survey Bulletin 1326, 83 p.

———, 1996, Pre-1980 tephra-fall deposits erupted from Mount St. Helens, Washington: U.S.Geological Survey Professional Paper 1563, 99 p.

Murphy, L. M., and Brazee, R. J., 1964, Seismological investigations of the Hebgen Lakeearthquake, in The Hebgen Lake, Montana, earthquake of August 17, 1959: U.S. GeologicalSurvey Professional Paper 435-C, p. 13-17.

Nathenson, M., 2001, Probabilities of volcanic eruptions and application to the recent history ofMedicine Lake volcano, in Vecchia, A. V., ed., A unified approach to probabilistic riskassessments for earthquakes, floods, landslides, and volcanoes, November 16-17, 1999,Golden, Colo.: U.S. Geological Survey Open-file Report 01-324, p. 71-71.

Newhall, C. G., and Dzurisin, D., 1988, Historical unrest at large calderas of the world: U. S.Geological Survey Bulletin 1855, 1108 p.

Obradovich, J. D., 1992, Geochronology of the late Cenozoic volcanism of Yellowstone NationalPark and adjoining areas, Wyoming an Idaho: U.S. Geological Survey Open-file Report 92-408, 45 p.

Otis, R. M., Smith, R. B., and Wold, R. J., 1977, Surveys of Yellowstone Lake, Wyoming: Journalof Geophysical Research, v. 82, p. 3705-3717.

Paperiello, R., 1998, The explosive history of Wall Pool and Black Opal Pool in Biscuit Basin:Geyser Observation and Study Association Transactions, v. 6, p. 106-114.

Pelton, J. R., and Smith, R. B., 1979, Recent crustal uplift in Yellowstone National Park: Science,v. 206, p. 1179-1182.

———, 1982, Contemporary vertical surface displacements in Yellowstone National Park: Journalof Geophysical Research, v. 87, p. 2745-2761.

Pierce, K. L., Adams, K. D., and Sturchio, N. C., eds., 1991, Geologic setting of the Corwin springsKnown Geothermal Resources Area-Mammoth Hot Springs area in and adjacent toYellowstone National Park: Menlo Park, Calif., U.S. Geological Survey Water-ResourcesInvestigations Report 91-4052, C1-C37 p.

Pierce, K. L., Cannon, K. P., Meyer, G. A., Trebesch, M. J., and Watts, R. D., 2002, Post-glacialinflation-deflation cycles, tilting, and faulting in the Yellowstone caldera based onYellowstone Lake shorelines: U.S. Geological Survey Open-file Report 02-0142, 63 p.,http://pubs.usgs.gov/of/2002/ofr-02-0142/.

Pierce, K. L., and Morgan, L. A., 1992, The track of the Yellowstone hotspot: Volcanism, faulting,and uplift, in Link, P. K., Kuntz, M. A., and Platt, L. B., eds., Regional Geology of easternIdaho and western Wyoming: Geological Society of America Memoir 179, p. 1-53.

Pierce, K. L., Morgan, L. A., and Saltus, R. W., 2000, Yellowstone plume head: Postulated tectonicrelations to the Vancouver slab, continental boundaries, and climate: U.S. GeologicalSurvey Open-file Report 00498, 39 p., http://pubs.usgs.gov/of/2000/ofr-00-0498/.

Pitt, A. M., Weaver, C. S., and Spence, W., 1979, The Yellowstone Park earthquake of June 30,1975: Seismological Society of America Bulletin, v. 69, p. 187-205.

Puskas, C., Smith, R. B., Meertens, C. M., and Chang, W. L., 2007, Crustal deformation of theYellowstone-Snake River Plain Volcanic System: Campaign and continuous GPSobservations, 1987-2004: Journal of Geophysical Research, v. 112, no. B03401,doi:10.1029/2006JB004325.

43

Puskas, C. M., Smith, R. P., and Meertens, C. M., 2002, GPS-derived models of intraplatedeformation of the Yellowstone hotspot: American Geophysical Union Transactions (Eos),v. 83, no. 47 (Fall Meeting Supplement), Abstract S11A-1104.

Rampino, M. R., 1991, Volcanism, climatic change, and the geologic record, in Fisher, R. V., andSmith, G. A., eds., Sedimentation in volcanic settings: Tulsa, Okla., Society forSedimentary Geology (SEPM), p. 9-18.

Rampino, M. R., and Ambrose, S. H., 2000, Volcanic winter in the Garden of Eden: The Tobasupereruption and the late Pleistocene human population crash, in McCoy, F. W., andHeiken, G., eds., Volcanic hazards and disasters in human antiquity: Boulder, Colo.,Geological Society of America Special Paper 345, p. 71-82.

Rampino, M. R., Self, S., and Strothers, R. B., 1988, Volcanic winters: Annual Reviews of Earthand Planetary Sciences, v. 16, p. 73-99.

Richins, W. D., Pechmann, J. C., Smith, R. B., Langer, C. J., Goter, S. K., Zollweg, J. E., and King,J. J., 1987, 1983 Borah Peak, Idaho, earthquake and its aftershocks: Seismological Societyof America Bulletin, v. 77, p. 694-723.

Scott, B. J., and Cody, A. D., 2000, Response of the Rotorua geothermal system to exploitation andvarying management regimes: Geothermics, v. 29, p. 573-592.

Scott, W. E., Iverson, R. M., Vallance, J. W., and Hildreth, W., 1995, Volcano hazards in theMount Adams region, Washington: U.S. Geological Survey Open-File Report 95-492, 11 p.

Shanks, W. C., Alt, J. C., and Morgan, L. A., in review, Geochemistry of sublacustrinehydrothermal deposits in Yellowstone Lake: Hydrothermal reactions, stable isotopesystematics, sinter deposition, and spire growth, in Morgan, L. A., ed., Integratedgeoscience studies in the Greater Yellowstone Geoecosystem: U.S. Geological SurveyProfessional Paper 1717.

Shanks, W. C., Morgan, L. A., Balistrieri, L., and Alt, J. C., 2005, Hydrothermal vent fluids,siliceous hydrothermal deposits, and hydrothermally altered sediments in YellowstoneLake, in Inskeep, W. P., and McDermott, T. R., eds., Geothermal biology and geochemistryin Yellowstone National Park: Bozeman, Mont., Montana State University Thermal BiologyInstitute, p. 53-72.

Sigurdsson, H., 1988, Gas bursts from Cameroon crater lakes: A new natural hazard: Disasters, v.12, p. 131-146.

Simmons, S. F., Keywood, M., Scott, B. J., and Keam, R. F., 1993, Irreversible change of theRotomahana-Waimangu hydrothermal system (New Zealand) as a consequence of avolcanic eruption: Geology, v. 21, p. 643-646.

Smith, E., and Bennett, K., 2006, Panther Creek volcano: Yellowstone Science, v. 14, no. 1, Winter2006, p. 5-12.

Smith, R. B., and Arabasz, W. J., 1991, Seismicity of the Intermountain Seismic Belt, inSlemmons, D. B., Engdahl, E. R., Zoback, M. D., and Blackwell, D. D., eds., Neotectonicsof North America: Boulder, Colo., Geological Society of America Decade Map Volume 1,p. 185-228.

Smith, R. B., and Braile, L. W., 1993, Topographic signature, space-time evolution, and physicalproperties of the Yellowstone-Snake River Plain volcanic system: The Yellowstone hotspot,in Snoke, A. W., Steidtmann, J., and Roberts, S. M., eds., Geology of Wyoming: Laramie,Wyo., Geological Survey of Wyoming Memoir 5, p. 694-754.

———, 1994, The Yellowstone hotspot: Journal of Volcanology and Geothermal Research, v. 61,p. 121-187.

Smith, R. B., Richins, W. D., and Doser, D. I., 1985, The 1983 Borah Peak, Idaho, earthquake:Regional seismicity, kinematics of faulting, and tectonic mechanism, in Stein, R. S., and

44

Bucknam, R. C., eds., Proceedings of Workshop XXVIII on the Borah Peak, Idaho,earthquake: U.S. Geological Survey Open-file Report OF85-0290, p. 236-263.

Smith, R. B., and Sbar, M. L., 1974, Contemporary tectonics and seismicity of the Western UnitedStates, with emphasis on the Intermountain Seismic Belt: Geological Society of AmericaBulletin, v. 85, p. 1205-1218.

Smith, R. B., Shuey, R. T., Pelton, J. R., and Bailey, J. P., 1977, Yellowstone Hot Spot, crustalproperties from new earthquake and magnetic data: Journal of Geophysical Research, v. 82,p. 3665-3676.

Sorey, M. L., Farrar, C. D., and Evans, W. C., 1996, Invisible CO2 gas killing trees at MammothMountain, California: U.S. Geological Survey Fact Sheet 172-96, 2 p.

Spell, T. L., Smith, E. I., Nastanski, N. M., and Bennett, K., 2004, Establishment and evolution of anew silicic magma system north of Yellowstone caldera: Geochronology, geochemistry, andpetrographic relationships of extracaldera basalts and rhyolites in the Norris-Mammothcorridor: American Geophysical Union Transactions (Eos), v. 85, no. 47, Fall meetingsupplement, abstract V52B-08.

Symonds, R. B., Gerlach, T. M., and Reed, M. H., 2001, Magmatic gas scrubbing: Implications forvolcano monitoring: Journal of Volcanology and Geothermal Research, v. 108, no. 1-4, p.303-341.

Tauxe, L., Luskin, C., Selkin, P., Gans, P., and Calvert, A., 2004, Paleomagnetic results from theSnake River Plain: Contribution to the time-averaged field global database: Geochemistry,Geophysics, and Geosystems, v. 5, no. 8, Q08H13, doi:10.1029/2003GC000661.

Taylor, R. C., and Grigg, J. B., 1999, Potts Hot Spring Basin Survey 1997-1998, unpublishedmanuscript: Geyser Observation and Study Association Press, 370 p.

Thompson, R. S., Hostetler, S. W., Bartlein, P. J., and Anderson, K. H., 1998, A strategy forassessing potential future changes in climate, hydrology, and vegetation in the WesternUnited States: U. S. Geological Survey Circular 1153, 20 p.

Traphagen, F. W., 1904, Death Gulch: Science, v. 19 (new series), no. 485, p. 632-634.Trimble, A. B., and Smith, R. B., 1975, Seismicity and contemporary tectonics of the Hebgen

Lake-Yellowstone Park region: Journal of Geophysical Research, v. 80, p. 733-741.U.S. Geological Survey, 1964, The Hebgen Lake, Montana, earthquake of August 17, 1959: U.S.

Geological Survey Professional Paper 435, 242 p.Vazquez, J. A., and Reid, M. R., 2002, Time scales of magma storage and differentiation of

voluminous high-silica rhyolites at Yellowstone caldera, Wyoming: Contributions toMineralogy and Petrology, v. 144, p. 274-285.

Waite, G. P., Smith, R. B., and Allen, R. L., 2006, Vp and Vs structure of the Yellowstone hot spotfrom teleseismic tomography: Evidence for an upper mantle plume: Journal of GeophysicalResearch, v. 111, no. B04303, doi:10.1029/2005JB003867.

Waite, G. P., and Smith, R. B., 2002, Seismic evidence for fluid migration accompanyingsubsidence of the Yellowstone caldera: Journal of Geophysical Research, v. 107, no. B9,2177, doi:10.1029/2001JB000586.

———, 2004, Seismotectonics and stress field of the Yellowstone volcanic plateau fromearthquake first-motions and other indicators: Journal of Geophysical Research, v. 109, no.B02301, doi:10.1029/2003JB002675.

Watson, B. N., 1961, Effects of the August 17, 1959 earthquake and subsequent quaking upon thethermal features of Yellowstone National Park: Tucson, Ariz., University of Arizona, M.S.thesis, 83 p.

Weed, W. H., 1889, A deadly gas-spring in the Yellowstone Park: Science, v. 13, no. 315, p. 130-132.

45

Werner, C., and Brantley, S., 2003, CO2 emissions from the Yellowstone volcanic system:Geochemistry, Geophysics, and Geosystems, v. 4, no. doi:10.1029/2002GC000473, p. 27.

White, B. J. P., 2006, Seismicity, seismotectonics and preliminary earthquake hazard analysis ofthe Teton region, Wyoming: Salt Lake City, University of Utah, M. S. thesis, 279 p.

White, D. E., 1955, Violent mud-volcano eruption of Lake City hot springs, northeasternCalifornia: Geological Society of America Bulletin, v. 66, no. 9, p. 1109-1130.

White, D. E., Fournier, R. O., Muffler, L. J. P., and Truesdell, A. H., 1975, Physical results ofresearch drilling in thermal areas of Yellowstone National Park, Wyoming: U.S. GeologicalSurvey Professional Paper 892, 70 p.

White, D. E., Hutchinson, R. A., and Keith, T. E. C., 1988, The geology and remarkable thermalactivity of Norris Geyser Basin, Yellowstone National Park, Wyoming: U.S. GeologicalSurvey Professional Paper 1456, 84 p.

White, D. E., Muffler, L. J. P., and Truesdell, A. H., 1971, Vapor-dominated hydrothermal systemscompared with hot-water systems: Economic Geology, v. 66, p. 75-97.

Whittlesey, L., 1988, Yellowstone Place Names: Helena, Montana, Montana Historical Society,178 p.

———, 1990, Monarch of all these mighty wonders: Tourists and Yellowstone’s Excelsior Geyser,1881-1890: Montana, The Magazine of Western History, v. 4, no. 2, p. 2-15.

———, 1995, Death in Yellowstone: Lanham, Md., Roberts Rinehart Publishers, 276 p.Wicks, C., Jr., Thatcher, W., and Dzurisin, D., 1998, Migration of fluids beneath Yellowstone

caldera inferred from satellite radar interferometry: Science, v. 282, p. 458-462.Wicks, C. W., Thatcher, W., Dzurisin, D., and Svarc, J., 2006, Uplift, thermal unrest, and magma

intrusion at Yellowstone caldera, observed with InSAR: Nature, v. 440, p. 72-75.Wilson, C. J. N., and Hildreth, W., 1997, The Bishop Tuff: New insights from eruptive

stratigraphy: Journal of Geology, v. 105, p. 407-439.Wold, R. J., Mayhew, M. A., and Smith, R. B., 1977, Bathymetric and geophysical evidence for a

hydrothermal explosion crater in Mary Bay, Yellowstone Lake, Wyoming: Journal ofGeophysical Research, v. 82, p. 3733-3738.

Yellowstone Volcano Observatory, 2006, Volcano and earthquake monitoring plan for theYellowstone Volcano Observatory, 2006-2015, U.S. Geological Survey ScientificInvestigations Report 2006-5276, 13 p., http://pubs.usgs.gov/sir/2006/5276/.

Mammoth HotSprings

Mammoth HotSprings

HebgenLake

HebgenLake

CanyonVillageCanyonVillage

M O N TA N A

M O N TA N AW Y O M I N G

WY

OM

ING

I D A H O

IDA

HO

MIDWAY GEYSERBASINMIDWAY GEYSERBASIN

UPPER GEYSERBASINUPPER GEYSERBASIN

West YellowstoneWest Yellowstone

GardinerGardiner

AshtonAshton

JacksonLake

JacksonLake

PITCHSTONEPLATEAU

PITCHSTONEPLATEAU

CENTRALPLATEAUCENTRALPLATEAU

MADISONPLATEAUMADISONPLATEAU

Mammoth HotSprings

Mammoth HotSprings

HebgenLake

HebgenLake

M O N TA N A

M O N TA N AW Y O M I N G

WY

OM

ING

I D A H O

IDA

HO

NORRIS GEYSERBASIN

NORRIS GEYSERBASIN

YellowstoneLake

CanyonVillageCanyonVillage

MadisonJunctionMadisonJunction

West YellowstoneWest Yellowstone

GardinerGardiner

AshtonAshton

JacksonLake

JacksonLake

Figure 20. Indian Pond, just north of Yellowstone Lake, is about 500 meters in maximum diameter and was formed about 3000 years ago by a hydrothermal explosion (Morgan and others, in review). Photo by Jim Peaco, 2001 (NPS stock photo).

65

77

Table 1. Ages of volcanic units related to the third cycle of the Yellowstone Plateau volcanic field.

Stratigraphic Unit Subunit Age±1 (Ma)* Source**

Third-cycle intracaldera units

Central Plateau Member Pitchstone Plateau flow 0.0791±0.0105 12

Grants Pass flow 0.072±0.003 1

Solfatara Plateau flow 0.1028±0.0076 13

Hayden Valley flow 0.102±0.004 3

West Yellowstone flow 0.1140±0.0012 12

Trischman Knob dome

Douglas Knob domeBechler River flow 0.116±0.002 2

Summit Lake flow 0.124±0.010 13

Tuff of Cold Mountain Creek 0.143±0.005 4

Spring Creek flow

Nez Perce Creek flow 0.1483±0.0051 13

dSpruce Creek flow

Elephant Back flow 0.153±0.002 1

Aster Creek flow 0.155±0.003 1

Buffalo Lake flow 0.160±0.003 1

Mary Lake flow 0.165±0.004 1

West Thumb flow 0.173±0.011 13

Tuff of Bluff Point 0.1731±0.0049 13

Dry Creek flow 0.166±0.009 4

Mallard Lake Member Mallard Lake flow 0.164±0.014 13

Upper Basin Member Scaup Lake flow 0.257±0.013 12

Dunraven Road flow 0.486±0.042 8

Canyon flow 0.484±0.015 7

Tuff of Sulphur Creek 0.473±0.009 9

Tuff of Uncle Toms Trail

Biscuit Basin flow 0.516±0.007 7

Third-cycle extracaldera unitsCentral Plateau Member Moose Falls flow 0.0806±0.0046 5

Roaring Mountain Member Crystal Spring flow 0.080±0.002 1

Obsidian Cliff flow 0.106±0.001 10

Gibbon River flow 0.118±0.010 10

Obsidian Creek Member Gibbon Hill dome 0.134±0.003 10

Roaring Mountain Member Norris Basin flow*** 0.138±0.003 1

Swan Lake Flat Basalt Panther Creek volcano 0.174±0.046 11

Snake River Group

Gerrit Basalt Hatchery Butte flow 0.390±0.038 14

Obsidian Creek Member Paintpot Hill dome 0.208±0.005 10

Osprey Basalt Lamar River flow 0.221±0.041 1

Obsidian Creek Member Landmark dome 0.226±0.006 10

Grizzly Lake complex 0.263±0.003 4

78

Gardner River complex 0.301±0.003 10

Apollinaris Spring dome 0.316±0.002 10

Willow Park dome 0.326±0.002 10

Madison River Basalt Flow 0.358±0.016 11

Roaring Mountain Member Cougar Creek dome 0.358±0.002 10

Riverside flow 0.526±0.003 10

Madison River Basalt "Vent area" flow 0.530±0.060 11

Falls River BasaltSwan Lake Flat Basalt Tower Road shield 0.590±0.065 11

Basalt of Geode CreekBasalt of Mariposa Lake

Third-cycle precaldera unitsUndine Falls Basalt Upper flow 0.584±0.026 1

Mount Jackson Rhyolite Big Bear Lake flowMount Jackson Rhyolite Mount Haynes flow 0.609±0.006 1

Basalt of Warm River Warm River flow 0.759±0.052 1

Mount Jackson Rhyolite Harlequin Lake flow 0.839±0.008 2Lewis Canyon Rhyolite Lewis Canyon flow 0.853±0.007 2Mount Jackson Rhyolite Flat Mountain flow 0.929±0.034 2

Mount Jackson Rhyolite Wapiti Lake flow 1.17±0.01 2Mount Jackson Rhyolite Moose Creek Butte flow 1.22±0.01 2

The Narrows Upper flow 1.30±0.35 2

*Shown in known or assumed stratigraphic order. Ages shown are selected as the “best” ages available on the basis of analyticalprecision, method of analysis, and fit to stratigraphic order. Error limits shown are 1 , as provided by the source, based on precisionof the analytical data. In some instances, geologic constraints demonstrate that the actual accuracy of the ages shown may reflecterrors somewhat larger than the stated analytical precision.**Data sources: 1. K-Ar sanidine age determination (Obradovich, 1992)

2. Weighted mean of 2 K-Ar sanidine age determinations (Obradovich, 1992)3. Weighted mean of 3 K-Ar sanidine age determinations (Obradovich, 1992)4. 40Ar/39Ar total-fusion sanidine age determination (J. D. Obradovich, 1997, unpubl. data)5. Weighted mean of 3 40Ar/39Ar total-fusion sanidine age determinations (J. D. Obradovich, 1997, unpubl.

data)6. 40Ar/39Ar total-fusion sanidine age determination (Muffler and others, 1971; Gansecki and others, 1996)7. 40Ar/39Ar total-fusion sanidine + plagioclase age determination (Gansecki and others, 1996)8. 40Ar/39Ar total-fusion plagioclase age determination (Gansecki and others, 1996)9. Weighted mean of 2 40Ar/39Ar total-fusion sanidine + plagioclase age determinations (Gansecki and others,

1996)10. 40Ar/39Ar age determination (N. Nastanski and T. L. Spell, 2004, unpubl. data)11. 40Ar/39Ar age determination (Smith and Bennett, 2006, from unpubl. data by N. Nastanski and T. L. Spell,

2004)12. 40Ar/39Ar plateau age determination (A. T. Calvert, 2005, unpublished)13. 40Ar/39Ar isochron age determination (A. T. Calvert, 2005, unpublished)14. 40Ar/39Ar plateau age determination (Tauxe and others, 2004)

*** Identified in the original source as part of the Gibbon River flow.

79

Table 2. Age, area, and volume of late postcaldera lavas of the third cycle of the Yellowstone

Plateau volcanic field.

Lava flows shown in most probable order of age, as based on data shown in figure 9. Cumulative volume of all flowsexcludes the estimated volume of unobserved older flows.

*Volumes for the two pyroclastic units cannot be calculated from existing data; the volumes shown with asterisks arerough estimates based only on the extents of known outcrop distributions and the size of the West Thumb caldera,source of the tuff of Bluff Point, as noted in the text.

Name of flow Age ±1

(ka)

Mean GroupAge±1

(ka)

Area

(km2)

Volume

(km3)

Cum.Volume

(km3)

Group Cum.Volume

(km3)

Avg. Thickness

(m)Pitchstone Plateau 79±11 351 70 366 200Grants Pass 72±3 76±5 14 0.5 296 71 37Solfatara Plateau 103±8 123 7 295 57Hayden Valley 102±4 103±1 111 2 288 9 17West Yellowstone 114±1 389 41 286 105Trischman Knob 0.8 0.014 245 17Douglas Knob 0.3 0.010 245 32Bechler River 116±2 236 8 245 35Summit Lake 124±10 118±5 402 37 237 86 92Cold Mountain Creek 143±5 10* 200Spring Creek 115 7 190 63Nez Perce Creek 148±5 157 6 183 41Spruce Creek 95 3 176 32Elephant Back 153±2 325 25 173 77Aster Creek 155±3 150±5 332 10 148 62 31Buffalo Lake 160±3 552 54 138 97Mary Lake 165±4 33 2 84 46West Thumb 173±11 245 11 83 43Bluff Point 173±5 50* 72Dry Creek 166±9 178 9 22 53

Mallard Lake 164±14 225 13 13 138 56

Unobserved units 167±5 2985 256 256 86

80

Table 3. Large hydrothermal explosion craters in Yellowstone National Park (data from Muffler and others, 1971; Morgan and others,

in review).

Feature Max..length

Max..width*

Diameter(avg.) Area

Max.Rim

height

Max.waterdepth

Totaldepthrim tofloor

Geologic Setting Age (est.)

(m) (m) (km) (km2) (m) (m) (m) (ka)Norris-Mammoth corridor:

Roaring Mountain east cratercomplex 644 504 0.574 0.255 58 0 58

In Lava Creek Tuff (LCT), along Norris-Mammoth corridor <16

Roaring Mountain north crater 314 289 0.302 0.071 33 0 33In Lava Creek Tuff, along Norris-Mammothcorridor

<16

The Gap-Norris Geyser Basin 114 105 0.162 0.009 16 unknown 16In Lava Creek Tuff, along Norris-Mammothcorridor

<16

Horseshoe Hill explosion craterIn Lava Creek Tuff, along Norris-Mammothcorridor

Lower and Upper Geyser Basins:

Pocket Basin 758 418 0.588 0.249 41 0 41In basin of Quaternary sediments surrounded byrhyolite lavas <16

Rush Lake 342 240 0.291 0.064 23 unknown 23In basin of Quaternary sediments surrounded byrhyolite lavas <16

Twin Buttes 641 624 0.633 0.314 129 unknown 129Thermal kame, in basin surrounded by rhyolitelavas <16

West Thumb area:

Duck Lake 733 500 0.617 0.288 35 18 53 Edge of Dry Creek flow 4-6Evil Twin 553 544 0.549 0.236 27 42 27 High heat flow, edge of Aster Creek flow 4-6Northern and Central Yellowstone Lake area:

Turbid Lake 1685 1502 1.594 1.988 85 42 127In alluvium, along topographic marginYellowstone caldera 10.3

Indian Pond 495 418 0.457 0.163 11 27 38In alluvium, along Weasel Creek-Storm Pointlinear trend 3.0

Mary Bay 2400 2824 2.612 5.323 60 53 113 High heat flow region of Yellowstone Lake 13.6Elliott's crater 938 727 0.833 0.536 52 60 52 High heat flow, inside edge of lava flow 8.0

Frank Island crater 770 712 0.741 0.431 21 50 21Edge Aster Creek flow, topo. marginYellowstone caldera

Upper Pelican river:

Sulphur Hills crater 354 248 0.301 0.069 50 0 50LCT, along resurgence-related fault in the SourCreek dome <16

81

Northeast Caldera area:

Fern Lake 1095 540 0.818 0.464 38 8 46Edge Upper Basin Member near topo. marginYellowstone caldera <16

Hot Spring Basin Group 289 268 0.279 0.243 39 0 39LCT and thermal kame, along topo. marginYellowstone caldera <16

Joseph's Coat 397 341 0.369 0.106 35 0 35Thermal kame, edge Canyon Flow & topo.margin Yell. caldera <16

*Maximum width is measured perpendicular to maximum length.

82

Table 4. Some historic hydrothermal explosion craters in Yellowstone National Park.

Feature Thermal areaMaximum

length(m)

Maximumwidth

(m)Age

(estimated)Reference

Upper Geyser BasinLink Geyser Biscuit Basin n.d. n.d. 1957-1958 (Marler and White, 1975)Black Opal/Wall Pool Biscuit Basin 60 15 1918, 1925,

1934, 1953(Marler and White, 1975; Paperiello, 1998)

Sapphire Biscuit Basin 10 6 1959 (Marler and White, 1975)Biscuit (2 small geysers) Biscuit Basin n.d. n.d. 1963 (Marler and White, 1975)Seismic Geyser Cascade Group 12 1959-1971 (Marler and White, 1975)near Sponge geyser Geyser Hill n.d. n.d. 1969 (Marler and White, 1975)Plume Geyser Geyser Hill n.d. n.d. 1922, 1972 (Marler, 1973)Lower and Midway Geyser BasinsFCG-3 of Bryan (1995) Fairy Group 3 3 1980? (Bryan, 1995)e. shoulder of Spasm geyser Fountain Group 1.5 1.2 1969 (Marler and White, 1975)S. Side White Creek Fountain Lake n.d. n.d. 1959 (Marler and White, 1975)Blowout Spring Kaleidoscope Group 4.5 4.5 1959 (Marler, 1973)18 m. north Kaleidoscope geyser Kaleidoscope Group 3 3 1963 (Marler and White, 1975)E. Shoulder of Honeycomb Kaleidoscope Group 3.7 1.8 1960 (Marler and White, 1975)West Flood Geyser Midway Geyser Basin 10.5 9 btw 1904-1940 (Marler and White, 1975)Excelsior Midway Geyser Basin 107 55 1881, 1882,

1888, 1890(Whittlesey, 1990)

Norris Geyser BasinPorcelain Terrace Norris Geyser Basin 9 2 1971 (White and others, 1988)Nymph Lake Norris Geyser Basin 75 21 2003 http://volcanoes.usgs.gov/yvo/2003/NorrisTherm03.htmlPorkchop Norris Geyser Basin 13.9 11.7 1989 (Fournier and others, 1991)Area 3 km northwest of Norris Unnamed 16 11 1986, 1987 (Hutchinson, 1987; Hobart, 1989)Unnamed Thermal area west of Elk

Parkn.d. n.d. 1988 (Paperiello, 1998)

Other AreasUnnamed Mushpots 1.9 1.4 1985 (Hutchinson, 1990)Unnamed vent 1 km S of LoneStar

Divide Group 1 1 1999 or 2000 Jeff Cross (personal commun., Feb. 2004)

Blowout Pool Shoshone Geyser Basin 5 5 1929 (Davis, 1930)Explosion Pool Potts Basin n.d. n.d. 1965, 1982 (Taylor and Grigg, 1999)60 North Potts Basin n.d. n.d. 1998 (Taylor and Grigg, 1999)Semicentennial Roaring Mountain n.d. n.d. 1922 (Whittlesey, 1988)

n.d. No data

83

Table 5. Estimates of annualized probability of events greater than a given magnitude.

Diameter (m) Area (m2) Events in last 14thousand years

AnnualizedProbability

>2 3.1 7000 (estimated) 0.50>300 70,700 16 0.0013

>2000 3,140,000 2 0.00014

84

Table 6. Ages of late intracaldera lavas and tuffs. Group ages for episodes calculated from the

means and standard deviations of the nominal ages. Model I arbitrarily spreads individual events

for each episode over 1000-year time intervals. Model II spreads individual events over plus and

minus one standard deviation for the three oldest episodes and, for the two youngest episodes

(each with only two events), at the event age and one standard deviation. Model III is an estimated

chronology based as closely as possible on the nominal ages and analytical standard deviations of

the individual events.

No. Name Ages and sd (ka)Group mean age

and sd (ka)Model I age

(ka)Model II age

(ka)Model III age

(ka)1 Pitchstone Plateau 79±11 72 72 722 Grants Pass 72±3

76±573 76 73

3 Solfatara Plateau 103±8 102 102 1024 Hayden Valley 102±4

103±1103 107 103

5 West Yellowstone 114±1 113.5 112 114

6 Trischman Knob 113.8 113 1167 Douglas Knob 114 114 116.2

8 Bechler River 116±2 114.3 115 116.49 Summit Lake 124±10

118±5

114.5 116 124

10 Cold Mountain Creek 143±5 151.5 149 14311 Spring Creek 151.7 150 145

12 Nez Perce Creek 148±5 151.9 152 14813 Spruce Creek 152.1 153 150

14 Elephant Back 153±2 152.3 154 15315 Aster Creek 155±3

150±5

152.5 155 155

16 Buffalo Lake 160±3 163.5 159 16017 Mary Lake 165±4 163.7 161 165

18 West Thumb 173±11 163.9 163 17219 Bluff Point 173±5 164.1 165 173

20 Dry Creek 166±9 164.3 167 17421 Mallard Lake 164±14

167±5

164.5 169 175

22 Unobserved units 165 171 176

85

Table 7. Conditional probabilities for an eruption in the next year for Models I, II, and III for the

mixed-exponential and Weibull distributions.

Mixed-exponential distribution

Time since lasteruption (y) 0 1000 20000 72000

Model

I 2.5x10-3 5.6x10-4 4.6x10-5 4.6x10-5

II 4.3x10-4 4.1x10-4 3.5x10-5 3.4x10-5

III 3.5x10-4 3.3x10-4 5.3x10-5 5.1x10-5

Weibull distribution

Time since lasteruption (y) 0 1000 20000 72000

Model

I 2.5x10-2 3.5x10-4 7.5x10-5 3.9x10-5

II 3.3x10-4 2.4x10-4 2.2x10-4 2.1x10-4

III 4.7x10-4 2.4x10-4 1.9x10-4 1.7x10-4

86

Appendices

Appendix 1. Description of representative historic hydrothermalexplosions

Porkchop Spring/GeyserThe most carefully documented hydrothermal explosion at Yellowstone occurred in 1989 at

Porkchop Geyser in Norris Geyser Basin. The following description is summarized from Fournierand others (1991).

Prior to 1985, Porkchop Geyser was a relatively obscure 3-m-long pool with a triangularorifice at the surface, a few cm on each side. In 1985, a change occurred wherein the pool drainedand was replaced by a 6- to 9-m-high “perpetual spouter” with continual pulsing of fluid. For fouryears, such behavior continued, with sufficient discharge to produce a roar audible up to 2 km awayfrom the vent. At 14:40 MDT on September 5, 1989, eight park visitors witnessed a tripling of theeruption column height, immediately followed by explosive disruption and ejection of the sintersheet surrounding the pool; the explosion was over in seconds. It resulted in formation of an 11-m-wide pool surrounded by an ejecta ring of sinter that was between 1 and 1.4 meters high. Materialwas ejected as far away from the pool as 66 meters to the south. The debris was not dispersedevenly as materials traveled only about 25 meters eastward. The maximum diameter of the largestejected block measured 1.9 meters. Fournier and others (1991) interpreted breccia blocks in theejected sinter as evidence of earlier explosions. In addition, a fluted fragment appeared to be part ofthe throat (or conduit) of the geyser. The throat was at least 10 cm in diameter, considerablygreater than the size of the orifice through which Porkchop had been venting to the surface.

Excelsior GeyserThe most witnessed, and most impressive of all historic hydrothermal explosions at

Yellowstone took place over about ten years between 1881 and 1890 at Excelsior Geyser in theMidway Geyser Basin. These explosions highlight the continuum between geyser eruptions andhydrothermal explosions. In essence, Excelsior exhibited geysering that was sufficiently intensethat rock fragments were commonly ejected and the size of the crater was markedly enlarged. Onehundred years after the cessation of its phenomenal geysering, Lee Whittlesey, the YellowstoneNational Park historian and archivist summarized the eyewitness accounts of the Excelsioreruptions (Whittlesey, 1990). The following material summarizes that information.

Excelsior was already a large and active 60 x 100 meter spring when the park wasdiscovered in 1872. Though eruptive activity may have started earlier, it was first described in1881, jetting to heights of 30 to 90 meters and, according to Superintendent P. W. Norris,sufficiently exceeding the activity of any other geyser that it merited the title of Excelsior. Activitycontinued into 1882, with occasional eruptions over 110 meters in height. Whittlesey (1990)quotes tour guide Nestor Henderson:

“The close of each eruption was accompanied by violent earthquake shocks that tore down thegeyserite walls and added much both to the danger and [to the] sublimity of the spectacle.These masses of broken wall were at each eruption hurled into the air several hundred feetabove the topmost waves, clashing together in their descent into the yawning abyss with adeafening noise that was most terrific.”

87

By the Fall of 1882, Excelsior became dormant. It may have erupted between 1883 and1888, but is clearly documented to have returned to activity by May 1, 1888, when blocks 0.3 m indiameter were hurled over 150 m from the crater. Eruptive activity continued throughout thesummer and fall. The last known eruption is believed to have been in 1890, until Excelsior eruptedagain for two days in 1985, over one hundred years after its first described eruptions.

Water discharge from Excelsior’s larger eruptions is said to have doubled the outflow of theFirehole River. The plume from Excelsior was visible at distances up to 160 km because the steamfrom its eruptions would create its own clouds.

West Nymph Creek Thermal AreaThe West Nymph Creek Thermal Area is an informal name for the unnamed backcountry

thermal area approximately 3 km north-northwest of the Norris Geyser Basin. The area was thesite of a series of hydrothermal explosions during 1987. In June, some time after activity hadstarted, two park employees witnessed a vent ejecting mud as high as six meters and as far away asten meters from the crater rim. As discussed in Hutchinson (1987), bursts from the pool were“churning freshly-killed tree trunks as if they were oversized swizzle sticks.” The crater size was9.1 x 10.5 m, with a depth as great as 3.45 m. This crater was 27 meters downslope from anothercrater inferred to have formed on January 26, 1986 as an effect of the autumn 1985 seismic swarm,in the northwest part of the Yellowstone caldera. Several other older but similar craters are foundwithin this thermal area. A report from a snow-coach driver combined with on-site observationsled the park geologist to conclude that the 1986 event was over in a matter of a few hours.

Black Opal/Wall Pool and Sapphire PoolExplosions that created the present Black Opal Pool and Wall Pool in the Upper Geyser

Basin may have begun as early as 1902, but are clearly documented in 1918, 1925, 1934 and 1953.By 1934, the opening of what is today “Wall Pool” was essentially complete. Subsequent activitywas entirely at Black Opal Pool. In January of 1934, a violent explosion created a pool 12 m indiameter that had not existed a week before. Several tons of rock were displaced, and some 300 kgrocks had been transported 9 m (Paperiello, 1998).

As noted above, explosive activity occurred in other years as well, and into the 1950s.After the August 17, 1959 Hebgen Lake earthquake, Black Opal Pool remained quiet, but nearbySapphire Pool exhibited impressive changes. For about three weeks after the earthquake, it surgedconstantly to a height of 3-4 meters. On September 5, massive bursts up to ~ 50m from an areaabout 60 m across began a period of activity that washed away the famous silica biscuits for whichBiscuit Basin was named (Marler, 1964).

Historic hydrothermal explosions elsewhereOn April 19, 2005, a 50m-wide crater resulted from a hydrothermal eruption in a remote

area of New Zealand between Taupo and Rotorua. Mud and water were dispersed 70-100 metersaway from the site, and the eruptive column, which reached 200 meters into the air, was visiblefrom 10 km away. The eruption was one of many that have occurred over the past 40 years in NewZealand. In 2000-2005 alone, eight different events were reported for New Zealand in the Bulletinof the Global Volcanism Network (http://www.volcano.si.edu). Scott and Cody (2000) reportedthat 91 hydrothermal explosions of natural features had occurred in Rotorua city between 1845 and1998, with the majority occurring at times of greatest human impact on the local hydrology andgeothermal systems.

In October 1990, a thermal area near El Barro, El Salvador exploded without apparentwarning, killing 25 people and injuring 15. The area was within a small community adjacent to the

88

producing Ahuachapán geothermal field. The eruption lasted a few minutes, created a crater 40 min diameter and dispersed rock fragments over a 200-m-diameter area (Handal and Barrios, 2004).Historical and geological studies indicate that the area had experienced previous hydrothermalexplosions.

Appendix 2. Description of large prehistoric hydrothermal eruption sitesat Yellowstone

Pocket BasinMuffler and others (1971) were the first to describe hydrothermal-explosion craters in

Yellowstone National Park, focusing primarily on Pocket Basin, a late Pleistocene ring offragmental deposits within the Lower Geyser Basin. Their study describes Pocket Basin as a flat-floored basin surrounded by an elliptical ridge of rock 360 x 790 m in minimum and maximumdiameters. The height of the ridge above the basin floor ranges from 4 to 20 m, with the innerslopes as steep as 20 to 25° and the shallower outer slopes less than 10°. The ridge is composed ofunconsolidated and unsorted materials that are predominantly angular fragments of yellow-stainedsandstone, siltstone and conglomerates. Most blocks in the deposit are less than 0.3 m in diameter,though a few range up to 2.5 m. The deposits are apparently only from shallow sources as onlyclasts from glacial deposits are present, with none of the underlying rhyolitic lava that wasencountered at depths of 40 meters in a nearby drillhole (Y3). Ejecta fragments are found as far as1200 meters from the center of Pocket Basin, and thus much farther than the actual berm oferuption deposits. The presence of glacial ice in the adjacent basin was inferred by Muffler andothers (1971) as an explanation for incision of the Firehole River through the Pocket Basin depositsrather than adjacent lowlands. This helps constrain the time of eruption to that of the latestglaciation, which ended ~16 ka. These authors speculated that catastrophic release of water from aglacial lake overlying areas within the Lower Geyser Basin may have caused an abrupt pressuredecrease within the geothermal system, thus triggering the eruption.

Mary BayIn the northern basin of Yellowstone Lake, Mary Bay contains an approximately 2.4-km by

2.8-km area of coalesced explosion craters (Wold and others, 1977), making it the world’s largestknown hydrothermal-explosion crater (Browne and Lawless, 2001). The Mary Bay hydrothermal-explosion crater extends from the northern basin of Yellowstone Lake onto land and into the lowerPelican Valley where steep (~35o slope) cliffs of explosion breccia are exposed. In the lake, a well-defined crater rim on the lake floor rises about 10 m above the flat-bottomed crater; the totalelevation difference between the flat-bottomed main crater floor to the top of the outer, subaerialcrater rim is as much as 70 m (Morgan and others, 2003b; Morgan and others, in review). About100 individual hydrothermal-vent structures are located within the larger main crater; individualvents are as deep as 35 m below the main crater floor, making the total elevation difference frombottom of an individual vent to the top of the rim close to 105 m. Investigations using asubmersible Remotely Operated Vehicle (ROV) show that fluids from a narrow, 53-m-deephydrothermal vent in Mary Bay have temperatures near the 120°C limit of the temperature probesused, reflecting extremely high heat-flow values in this area (Morgan and others, 1977). Charcoalfragments collected from the soil layer below the Mary Bay ejecta deposits as well as from withinthe hydrothermal explosion deposit have an average calibrated 14C age of 13.6 ka BP (Pierce andothers, 2002).

89

Elliott’s craterWithin Yellowstone Lake, 1 kilometer southwest of the Mary Bay crater complex is a large

(~800-m-diameter) composite depression informally referred to as Elliott’s crater (Morgan andothers, 2003b), named after Henry Elliott, who mapped Yellowstone Lake during the 1871 Haydensurvey. The northern rim of the crater rises about 40 m above the flat-bottomed main crater floor,whereas the southern crater rim is about 30 m above the crater floor. About 10 smaller individualcraters are present on the main crater floor and along the southwest rim; several of these are activeand have fluid temperatures ranging from 51ºC to as high as 91ºC. The presence of two youngercraters at the south end of the main crater floor and the presence of many smaller hydrothermalvents further indicates more recent hydrothermal activity and possibly younger explosions. Alongthe southeast and eastern rim outside of the main crater, more than 3 km south from shore, rocks ofvaried lithologic compositions lie above sediments on the lake floor; these rocks may be productsof more recent hydrothermal explosions.

Stratigraphic relations indicate that the main hydrothermal explosion occurred between 13and 8 ka, based on sedimentation rates in the lake (Johnson and others, 2003). Seismic reflectiontechniques identify “opaque areas” within the sedimentary fill that are interpreted to indicate thepresence of hydrothermal fluids and (or) gases (Johnson and others, 2003; Morgan and others,2003b).

Evil Twin explosion craterMorgan and others (2003b) described a 500-m-diameter, sublacustrine explosion crater,

referred to informally as the Evil Twin explosion crater, in the western part of West Thumb basinnear the currently active West Thumb Geyser Basin and 300 m northeast of Duck Lake, apostglacial (subaerial) hydrothermal-explosion crater just outside Yellowstone Lake (Muffler andothers, 1971). The 500-m-wide Evil Twin explosion crater is surrounded by 12- to 20-m-high,nearly vertical, walls and has several smaller nested craters along its eastern edge. These nestedcraters are as deep as 40 m and are younger than the main crater. Temperatures of hydrothermalfluids emanating from the smaller northeastern nested crater have been measured at 72°C by ROV(Morgan and others, in review).

Frank Island explosion craterAn oval, steep-walled, flat-bottomed hydrothermal-explosion crater more than 700 m wide

is located south of Frank Island (Morgan and others, 2003b). Previous lower-resolution seismicreflection profiles interpreted this structure as a topographic edge to the Yellowstone caldera (Otisand others, 1977); however, higher-resolution swath sonar and shallower seismic-reflection profilesindicate that this structure, while on the slumped margin of the Yellowstone caldera, hascharacteristics very similar to those of other hydrothermal explosion craters in Yellowstone Lake(Morgan and others, 2003b). Muted topography suggests that this explosion crater is one of theoldest still recognizable in Yellowstone Lake. Further, this crater occurs in an area where heat-flowvalues are, at present, relatively low. Submersible ROV investigations do not indicate hydrothermalactivity within the crater (Morgan and others, in review).

Indian PondIndian Pond, formerly referred to as Squaw Lake (Muffler and others, 1971), is an oval,

500-m-wide, lake-filled crater (Table 3; Figure 20). The pond is rimmed by an apron of explosionbreccia that rises about 11 m above present-day lake level. Low-resolution bathymetric surveys ofIndian Pond (Unpubl. Data, Yellowstone National Park, 1966) indicate the pond has steep inward-

90

dipping slopes around its perimeter and has at least three elongate, east-west-trending basins on thecrater floor. The average depth of the crater floor is about 13 m below lake level; depths of theindividual basins or smaller craters range from 20 to 27 m below lake level. The lake is breachedon its southwest edge where Little Indian Creek flows out of Indian Pond into northernYellowstone Lake.

The Indian Pond explosion breccia is a poorly sorted, matrix-supported breccia deposit; thematrix is light medium-brown clay that has a pervasive orange-yellow stain where groundwater hasflowed. Lithic clasts in the breccia are generally angular to sub-angular and are composedprimarily of cemented beach gravels with subordinate angular clasts of silicified lake sedimentswith hydrothermal clinoptilolite and well-sorted, fine-grained indurated golden-tan siltstone.Maximum diameter lithic fragments of cemented beach gravels and sands are up to 1.5 m at thecrater rim; in the wave-cut terrace exposures along Mary Bay, lithic fragments have maximumdiameters of 30 cm and average around 3 to 7 cm. The thickness of the Indian Pond breccia in thewave cut terraces along Mary Bay is less than one meter, although northwest of the crater,thicknesses are estimated to be as much as several meters.

Turbid LakeTurbid Lake, along the eastern edge of the topographic margin of the Yellowstone caldera

was recognized by Muffler and others (1971) as a hydrothermal explosion crater. It is the secondlargest hydrothermal explosion crater in Yellowstone, having a maximum crater diameter of 1685meters and an area close to 2.0 km2 (Morgan and others, in review). A large primary crater islocated in the main central portion of the lake and is rimmed by an apron of explosion breccia thatrises about 33 m above present-day lake level on its northern, western, and southern shores. Alongits eastern edge, evidence for a smaller second crater is present where a north-south-trending ridgeof explosion breccia is deposited inside the main crater wall. The height of the eastern main craterwall rises about 85 m above present-day level of Turbid Lake. Low-resolution bathymetric surveysof Turbid Lake (Unpubl. data, Yellowstone National Park, 1977) indicate that the lake has steepinward-dipping slopes around its central deep crater and has multiple smaller craters around theperiphery of the deep crater. Maximum depth of the central deep crater is 42 m below lake level;depths of the smaller craters range from 17 m to 5 m along the northern and western edges to asdeep as 27 m along the eastern edge (table 3, Morgan and others, in review). The crater rim isbreached on the north by the Sedge Creek inlet and on the south by the inlet of Bear Creek. At itswestern crater rim, Sedge Creek flows out of Turbid Lake and into northern Yellowstone Lake.Acidic hot springs are present along the eastern and southeastern rim; a thermal area is presentalong the southern rim at lake level. Mud pots there have pH of 1.1-1.5 and temperatures of 49-57°C.

The hydrothermal breccia is exposed along the creeks, varying in thickness from 2 to 10 m.The matrix-supported breccia is composed of angular fragments of hydrothermally altered LavaCreek Tuff, cemented gravels and sands, moderately and poorly sorted clastic sulfidic sandstones,cemented pebble conglomerate, and chert breccias (Morgan and others, in review). The matrix isgenerally whitish fine-grained clay. As noted by Muffler and others (1971), a broad constructionalouter ramp extends from Turbid Lake toward the northwest and may be indicative of the primaryflow direction of the explosion deposit. The explosion breccia may be distributed toward thenorthwest and west as much as 4.5 km from its source. Charcoal fragments from within thehydrothermal explosion deposit from the southern rim have a radiometric age of 9.4 ka (Pierce andothers, 2002).

91

Appendix 3. Probabilities of episodic volcanic eruptions and applicationto the young intracaldera volcanic history of Yellowstone

By Manuel Nathenson and Robert L. Christiansen

An underlying assumption of USGS volcano hazards assessments for Cascade volcanoes inOregon and Washington has been that the probability distribution of volcanic eruptions may betreated as a Poisson process. Time histories for some volcanoes match this assumption well (e.g.Klein, 1982). The probability of an eruption during any particular period of time is calculated fromthe relation for the occurrence rate. For a Poisson process, this relation is obtained from theexponential distribution for the probability P{ T t} that an eruption will occur in a time T lessthan or equal to the time period t:

P{ T t} = F(t) = 1 - e-μ t

μt , for μt small,where F(t) is the probability distribution function, and μ is the mean occurrence rate (events

per year) for the exponential distribution. Investigators such as Scott and others (1995) cite lowoccurrence rates in the Cascades as justification for use of this relation there.

Given a set of n eruption time intervals ti, the average recurrence interval (the reciprocal ofthe occurrence rate) may be estimated by:

1μ = 1n

i=1

nt i

The properties of a Poisson process include the characteristic that the conditionalprobability of waiting a time until an eruption occurs does not depend on the time already waitedbut only on the time period selected (e.g. 1 year, 30 years, etc.) to calculate a conditionalprobability. For some volcanoes, the time history contains disparate time intervals betweeneruptions, some being short and others much longer. Some examples of time histories having suchdisparate eruption-time intervals are those of Mount Rainier and Mount St. Helens in Washington.Mullineaux’s (1974) data for tephra layers at Mount Rainier include three long intervals (>2000years) and seven short intervals (<600years) between eruptions. Mullineaux's (1996) data forMount St. Helens include one interval of 8600 years, one of 1500 years, and 34 less than 640 years.In such instances, other probability distributions more accurately represent the data, the conditionalprobabilities based on these distributions depending on the time since the last eruption.

Bebbington and Lai (1996) proposed using the Weibull distribution to model eruption timesthat vary with the preceding time interval:

P{ T t} = F(t) = 1- e-μ(t)

where μ(t) = (t )

and T is the time, less than the time period t, when an eruption will occur. Parameters and are referred to as the scale and shape parameters, respectively; when = 1, this reduces to the

exponential distribution.For eruption intervals that can be divided into two populations, one of short intervals and

one of long intervals, a relevant model is the mixed exponential (Cox and Lewis, 1966; Nathenson,2001):

P{ T t} = F(t) = 1- p1 e-μ1 t - p2 e-μ2 t

where p1 = n1

n1+ n2

92

and 1μ1

= 1n1

i=1

n1

t i

where p1 is the fraction of short intervals, μ1 is the mean occurrence rate for the shortintervals, n1 is the number of short intervals, and p2, μ2, and n2 are equivalent parameters for thelong intervals. The basic notion embodied in the mixed exponential is that there are two states.The probability of an eruption occurring in each of these states is governed by an exponentialdistribution. If one knew that the volcano were currently in a particular state (a difficult judgmentto make), then the probability of an eruption could be calculated using the appropriate simpleexponential relation for that state only.

Of direct interest for application here is the conditional probability P{ t T t + t | T >t} of an eruption occurring between time t and time t + t, (e.g. the next year or the next 30

years), after already waiting a time t since the last eruption. This conditional probability can becalculated from the distribution function F(t) as

P{ t T t + t | T > t} = 1 - 1- F(t+ t)

1- F( t)

For the simple exponential distribution, the conditional probability reduces to:P{ t T t + t | T > t} = 1 - e-μ t

Thus, for the simple exponential distribution, the passage of past time does not change theprobability of the time to a future eruption. (In the engineering language of time to failure, there isno wear or fatigue). For the Weibull distribution, the conditional probability is:

P{ t T t + t | T > t} = 1 – exp {(t ) - (

t + t ) }

For the mixed exponential, the conditional probability is:P{ t T t + t | T > t} = 1 – [p1 e-μ1 (t + t) + p2 e-μ2 (t + t) ] /

[p1 e-μ1 t + p2 e-μ2 t ]Thus, unlike for the simple exponential distribution, the conditional probability for the

mixed exponential does depend on the time since the last eruption, t.As discussed in the section on hazards from large rhyolitic lava eruptions, the chronology of

these eruptions is problematic. Table 6 reproduces the ages for these events as given in tables 1and 2. Some of these rhyolites yield nominal ages that are younger that those of rhyolites that arestratigraphically above them, but all such discrepancies with stratigraphic order remain within therange of analytical uncertainty, reported as plus or minus one standard deviation. The combinedrecord of mapped stratigraphy and isotopic ages suggests that these rhyolites represent a series ofepisodes. The times between episodes are tens of thousands of years, but the times between eventswithin each episode are short—possibly tens, hundreds or a few thousand years. With availableisotopic-age data, it is not possible to discern the actual time intervals between eruptions within anepisode, and there could be intervals in one episode that are a few tens of years and in anotherepisode that are hundreds or thousands of years. We use the available data to calculate both a meanage for each episode, weighted by the analytical uncertainty for each event within the episode, anda standard deviation for the age of the episode (table 6). The mean recurrence interval for the fiveepisodes is 23 ky, with a range from 12 ky to 38 ky; treating each episode as an independent event,the probability in one year of a new episode occurring is 4.3x10-5.

To evaluate the variation in eruption probabilities resulting from the uncertainty of thevarious time intervals between eruptions within an episode, we assess three models. Model Iassumes that each episode lasts a thousand years; the assumed individual eruption ages are spreadevenly within that thousand years for each episode, as listed in table 6. Model II assumes that the

93

calculated standard deviation for the mean age of each episode represents a measure, not only ofthe analytical uncertainty, but also the actual duration of the episode, spreading the individualevents evenly over the calculated standard deviation. For the two youngest episodes, each withonly two events, the age of one event in this model is taken to be the mean age, and the other istaken to be the mean age plus one standard deviation; for the other three episodes, the ages arespread over a time equal to plus and minus one standard deviation. Model III is an estimatedchronology based as closely as possible on the reported ages and standard deviations for eacheruptive event. When the nominal ages are in correct stratigraphic order, they are assigned directly.Other ages are assigned to be as close as possible to the nominal ages and both to be within thereported standard deviation and to comply with the proposed stratigraphic order. (For a few events,where stratigraphic order is ambiguous, as illustrated in figure 9, the order for model III is chosenrather than definitively established).

The three models tabulated in table 6 are illustrated in figure 27. Model I, with its shortduration for each episode, readily resolves graphically into the five episodes discussed here. ModelII, with its longer duration of each episode, still resolves graphically into the episodes but lessclearly than model I. The episodes are not well resolved graphically in model III. Some of theepisodes correspond graphically more to a change in rate of occurrence than to distinctly differenttimes. All three models, however, preserve the observation that there are long periods with nointracaldera rhyolitic eruptive activity.

Probability distributions calculated for the three models (fig. 28) all show that mostintervals are grouped at relatively short times and only a few at longer or much longer times.Model I (fig. 28A) has a large number of very short time intervals between eruptions and fourlonger intervals. Model II (fig. 28B) spreads the eruption time intervals more uniformly, with onlytwo long intervals. Model III (fig. 28C) spreads the time intervals more uniformly than either ofthe other two models. The three analytical distribution functions are also shown for each model.As expected, because of the disparate time intervals, the simple exponential distribution is a poor fitto the data in models I and II, but it is a reasonable fit to the model III data. The Weibulldistribution is a reasonably good fit to model I but a poor fit to both models II and III for the long-interval data. The mixed exponential distribution provides a good fit to both the short-interval andthe long-interval data for each model.

The conditional probability of an eruption occurring in the next year is shown in figure 29for each of the three models. Some representative values are listed in table 7, including some of theprobabilities at short times after an eruption that are off the scale of the plots in figure 29. Theconditional probability based on the simple exponential distribution does not depend on the timesince the last eruption. Because each of the three models has a somewhat different total duration,(table 6), the probabilities based on the exponential distribution vary slightly among the three partsof figure 29. The conditional probabilities based on the Weibull distribution for long times sincethe last eruption are considerably different for model I than for models II and III (fig. 29 and table7). The mixed exponential distribution yields similar results for long times since the last eruptionin all three models. Because of the poor match of the Weibull distribution for long time intervalswith models II and III (fig. 28) and the variable results for the conditional probability at long times(fig. 29 and table 7), the mixed exponential distribution is a better choice for the calculation ofconditional probabilities for this data set.

The greatest differences for conditional probabilities among the three models for the mixedexponential distribution are at short times. For the probability of an eruption in the next yearimmediately following an eruption, the mixed exponential distribution yields a probability almostten times higher for model I than for models II and III (table 7). This result accords with theintuitive recognition that the many short time intervals in model I would lead one to anticipateanother eruption shortly after one had occurred. At the present time, a value for the probability of

94

an intracaldera rhyolitic eruption within the next year of 5x10-5 reasonably represents all threemodels based on a mixed exponential distribution. This value is quite close to the value of 4x10-5

calculated independently for a new episode. Because both calculations are model dependent, theirvery similar results are reassuring; it has been ~72 ky since the last eruption, which is also the timesince the last episode. The major differences among the models are in the estimation of theprobabilities of another eruption to follow if an eruption were to occur now (table 7). Thelimitations of the available data preclude a definitive resolution of those differences and heightenthe uncertainty associated with their use in estimating eruption probabilities.

In addition to the probability of a future intracaldera rhyolitic eruption occurring within anygiven time period, an important consideration would be its likely size. Figure 30 illustrates theprobability of a future lava eruption within the caldera having a volume greater that any particularvalue, based upon volumes estimated for 20 of the known late intracaldera eruptions (91 % of theknown events), excluding the two pyroclastic eruptions, whose volumes are indeterminate (table 3).The data are distributed logarithmically with volume over most of the range. The probability that afuture intracaldera lava flow would have a volume greater than 2 km3 is about 80%; the probabilityof a volume greater than 10 km3 is about 40%. Thus, the probability of a large-volume future lavaflow in the caldera is greater than the probability of a small one.